Datasets:

WINGNUS
/

ACL-OCL

	{
	"paper_id": "J02-4002",
	"header": {
	"generated_with": "S2ORC 1.0.0",
	"date_generated": "2023-01-19T02:44:15.174566Z"
	},
	"title": "Summarizing Scientific Articles: Experiments with Relevance and Rhetorical Status",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": "",
	"affiliation": {},
	"email": "simone.teufel@cl.cam.ac.uk"
	},
	{
	"first": "Marc",
	"middle": [],
	"last": "Moens",
	"suffix": "",
	"affiliation": {},
	"email": ""
	}
	],
	"year": "",
	"venue": null,
	"identifiers": {},
	"abstract": "In this article we propose a strategy for the summarization of scientific articles that concentrates on the rhetorical status of statements in an article: Material for summaries is selected in such a way that summaries can highlight the new contribution of the source article and situate it with respect to earlier work. We provide a gold standard for summaries of this kind consisting of a substantial corpus of conference articles in computational linguistics annotated with human judgments of the rhetorical status and relevance of each sentence in the articles. We present several experiments measuring our judges' agreement on these annotations. We also present an algorithm that, on the basis of the annotated training material, selects content from unseen articles and classifies it into a fixed set of seven rhetorical categories. The output of this extraction and classification system can be viewed as a single-document summary in its own right; alternatively, it provides starting material for the generation of task-oriented and user-tailored summaries designed to give users an overview of a scientific field.",
	"pdf_parse": {
	"paper_id": "J02-4002",
	"_pdf_hash": "",
	"abstract": [
	{
	"text": "In this article we propose a strategy for the summarization of scientific articles that concentrates on the rhetorical status of statements in an article: Material for summaries is selected in such a way that summaries can highlight the new contribution of the source article and situate it with respect to earlier work. We provide a gold standard for summaries of this kind consisting of a substantial corpus of conference articles in computational linguistics annotated with human judgments of the rhetorical status and relevance of each sentence in the articles. We present several experiments measuring our judges' agreement on these annotations. We also present an algorithm that, on the basis of the annotated training material, selects content from unseen articles and classifies it into a fixed set of seven rhetorical categories. The output of this extraction and classification system can be viewed as a single-document summary in its own right; alternatively, it provides starting material for the generation of task-oriented and user-tailored summaries designed to give users an overview of a scientific field.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Abstract",
	"sec_num": null
	}
	],
	"body_text": [
	{
	"text": "Summarization systems are often two-phased, consisting of a content selection step followed by a regeneration step. In the first step, text fragments (sentences or clauses) are assigned a score that reflects how important or contentful they are. The highestranking material can then be extracted and displayed verbatim as \"extracts\" (Luhn 1958; Edmundson 1969; Paice 1990; Kupiec, Pedersen, and Chen 1995) . Extracts are often useful in an information retrieval environment since they give users an idea as to what the source document is about (Tombros and Sanderson 1998; , but they are texts of relatively low quality. Because of this, it is generally accepted that some kind of postprocessing should be performed to improve the final result, by shortening, fusing, or otherwise revising the material (Grefenstette 1998; Mani, Gates, and Bloedorn 1999; Jing and McKeown 2000; Barzilay et al. 2000; Knight and Marcu 2000) .",
	"cite_spans": [
	{
	"start": 333,
	"end": 344,
	"text": "(Luhn 1958;",
	"ref_id": "BIBREF49"
	},
	{
	"start": 345,
	"end": 360,
	"text": "Edmundson 1969;",
	"ref_id": "BIBREF29"
	},
	{
	"start": 361,
	"end": 372,
	"text": "Paice 1990;",
	"ref_id": "BIBREF59"
	},
	{
	"start": 373,
	"end": 405,
	"text": "Kupiec, Pedersen, and Chen 1995)",
	"ref_id": "BIBREF42"
	},
	{
	"start": 544,
	"end": 572,
	"text": "(Tombros and Sanderson 1998;",
	"ref_id": "BIBREF81"
	},
	{
	"start": 803,
	"end": 822,
	"text": "(Grefenstette 1998;",
	"ref_id": "BIBREF32"
	},
	{
	"start": 823,
	"end": 854,
	"text": "Mani, Gates, and Bloedorn 1999;",
	"ref_id": "BIBREF51"
	},
	{
	"start": 855,
	"end": 877,
	"text": "Jing and McKeown 2000;",
	"ref_id": "BIBREF37"
	},
	{
	"start": 878,
	"end": 899,
	"text": "Barzilay et al. 2000;",
	"ref_id": null
	},
	{
	"start": 900,
	"end": 922,
	"text": "Knight and Marcu 2000)",
	"ref_id": "BIBREF40"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "The extent to which it is possible to do postprocessing is limited, however, by the fact that contentful material is extracted without information about the general discourse context in which the material occurred in the source text. For instance, a sentence describing the solution to a scientific problem might give the main contri-bution of the paper, but it might also refer to a previous approach that the authors criticize. Depending on its rhetorical context, the same sentence should be treated very differently in a summary. We propose in this article a method for sentence and content selection from source texts that adds context in the form of information about the rhetorical role the extracted material plays in the source text. This added contextual information can then be used to make the end product more informative and more valuable than sentence extracts.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "Our application domain is the summarization of scientific articles. Summarization of such texts requires a different approach from, for example, that used in the summarization of news articles. For example, Barzilay, McKeown, and Elhadad (1999) introduce the concept of information fusion, which is based on the identification of recurrent descriptions of the same events in news articles. This approach works well because in the news domain, newsworthy events are frequently repeated over a short period of time. In scientific writing, however, similar \"events\" are rare: The main focus is on new scientific ideas, whose main characteristic is their uniqueness and difference from previous ideas.",
	"cite_spans": [
	{
	"start": 207,
	"end": 244,
	"text": "Barzilay, McKeown, and Elhadad (1999)",
	"ref_id": "BIBREF22"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "Other approaches to the summarization of news articles make use of the typical journalistic writing style, for example, the fact that the most newsworthy information comes first; as a result, the first few sentences of a news article are good candidates for a summary (Brandow, Mitze, and Rau 1995; Lin and Hovy 1997) . The structure of scientific articles does not reflect relevance this explicitly. Instead, the introduction often starts with general statements about the importance of the topic and its history in the field; the actual contribution of the paper itself is often given much later.",
	"cite_spans": [
	{
	"start": 268,
	"end": 298,
	"text": "(Brandow, Mitze, and Rau 1995;",
	"ref_id": "BIBREF24"
	},
	{
	"start": 299,
	"end": 317,
	"text": "Lin and Hovy 1997)",
	"ref_id": "BIBREF48"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "The length of scientific articles presents another problem. Let us assume that our overall summarization strategy is first to select relevant sentences or concepts, and then to synthesize summaries using this material. For a typical 10-to 20-sentence news wire story, a compression to 20% or 30% of the source provides a reasonable input set for the second step. The extracted sentences are still thematically connected, and concepts in the sentences are not taken completely out of context. In scientific articles, however, the compression rates have to be much higher: Shortening a 20-page journal article to a half-page summary requires a compression to 2.5% of the original. Here, the problematic fact that sentence selection is context insensitive does make a qualitative difference. If only one sentence per two pages is selected, all information about how the extracted sentences and their concepts relate to each other is lost; without additional information, it is difficult to use the selected sentences as input to the second stage.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "We present an approach to summarizing scientific articles that is based on the idea of restoring the discourse context of extracted material by adding the rhetorical status to each sentence in a document. The innovation of our approach is that it defines principles for content selection specifically for scientific articles and that it combines sentence extraction with robust discourse analysis. The output of our system is a list of extracted sentences along with their rhetorical status (e.g. sentence 11 describes the scientific goal of the paper, and sentence 9 criticizes previous work), as illustrated in Figure 1 . (The example paper we use throughout the article is F. Pereira, N. Tishby, and L. Lee's \"Distributional Clustering of English Words\" [ACL-1993, cmp lg/9408011] ; it was chosen because it is the paper most often cited within our collection.) Such lists serve two purposes: in themselves, they already provide a better characterization of scientific articles than sentence extracts do, and in the longer run, they will serve as better input material for further processing.",
	"cite_spans": [
	{
	"start": 757,
	"end": 783,
	"text": "[ACL-1993, cmp lg/9408011]",
	"ref_id": null
	}
	],
	"ref_spans": [
	{
	"start": 613,
	"end": 621,
	"text": "Figure 1",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "An extrinsic evaluation (Teufel 2001) shows that the output of our system is already a useful document surrogate in its own right. But postprocessing could turn Summarizing Scientific Articles AIM 10 Our research addresses some of the same questions and uses similar raw data, but we investigate how to factor word association tendencies into associations of words to certain hidden senses classes and associations between the classes themselves. 11 While it may be worthwhile to base such a model on preexisting sense classes (Resnik, 1992) , in the work described here we look at how to derive the classes directly from distributional data. 162 We have demonstrated that a general divisive clustering procedure for probability distributions can be used to group words according to their participation in particular grammatical relations with other words.",
	"cite_spans": [
	{
	"start": 24,
	"end": 37,
	"text": "(Teufel 2001)",
	"ref_id": "BIBREF76"
	},
	{
	"start": 527,
	"end": 541,
	"text": "(Resnik, 1992)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1."
	},
	{
	"text": "19 The corpus used in our first experiment was derived from newswire text automatically parsed by Hindle's parser Fidditch (Hindle, 1993) . 113 The analogy with statistical mechanics suggests a deterministic annealing procedure for clustering (Rose et al., 1990) , in which the number of clusters is determined through a sequence of phase transitions by continuously increasing the parameter EQN following an annealing schedule. CONTRAST 9 His notion of similarity seems to agree with our intuitions in many cases, but it is not clear how it can be used directly to construct word classes and corresponding models of association. 14 Class construction is then combinatorially very demanding and depends on frequency counts for joint events involving particular words, a potentially unreliable source of information as we noted above.",
	"cite_spans": [
	{
	"start": 98,
	"end": 137,
	"text": "Hindle's parser Fidditch (Hindle, 1993)",
	"ref_id": null
	},
	{
	"start": 243,
	"end": 262,
	"text": "(Rose et al., 1990)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "BASIS",
	"sec_num": null
	},
	{
	"text": "Extract of system output for example paper.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 1",
	"sec_num": null
	},
	{
	"text": "0 This paper's topic is to automatically classify words according to their contexts of use. 4 The problem is that for large enough corpora the number of possible joint events is much larger than the number of event occurrences in the corpus, so many events are seen rarely or never, making their frequency counts unreliable estimates of their probabilities. 162 This paper's specific goal is to group words according to their participation in particular grammatical relations with other words, 22 more specifically to classify nouns according to their distribution as direct objects of verbs.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 1",
	"sec_num": null
	},
	{
	"text": "Nonexpert summary, general purpose.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "the rhetorical extracts into something even more valuable: The added rhetorical context allows for the creation of a new kind of summary. Consider, for instance, the user-oriented and task-tailored summaries shown in Figures 2 and 3. Their composition was guided by fixed building plans for different tasks and different user models, whereby the building blocks are defined as sentences of a specific rhetorical status. In our example, most textual material is extracted verbatim (additional material is underlined in Figures 2 and 3 ; the original sentences are given in Figure 5 ). The first example is a short abstract generated for a nonexpert user and for general information; its first two sentences give background information about the problem tackled. The second abstract is aimed at an expert; therefore, no background is given, and instead differences between this approach and similar ones are described. The actual construction of these summaries is a complex process involving tasks such as sentence planning, lexical choice and syntactic realization, tasks that are outside the scope of this article. The important point is that it is the knowledge about the rhetorical status of the sentences that enables the tailoring of the summaries according to users' expertise and task. The rhetorical status allows for other kinds of applications too: Several articles can be summarized together, contrasts or complementarity among 44 This paper's goal is to organise a set of linguistic objects such as words according to the contexts in which they occur, for instance grammatical constructions or n-grams. 22 More specifically: the goal is to classify nouns according to their distribution as direct objects of verbs. 5 Unlike Hindle (1990) , 9 this approach constructs word classes and corresponding models of association directly. 14 In comparison to Brown et al. (1992) , the method is combinatorially less demanding and does not depend on frequency counts for joint events involving particular words, a potentially unreliable source of information. articles can be expressed, and summaries can be displayed together with citation links to help users navigate several related papers.",
	"cite_spans": [
	{
	"start": 1736,
	"end": 1749,
	"text": "Hindle (1990)",
	"ref_id": null
	},
	{
	"start": 1862,
	"end": 1881,
	"text": "Brown et al. (1992)",
	"ref_id": null
	}
	],
	"ref_spans": [
	{
	"start": 518,
	"end": 533,
	"text": "Figures 2 and 3",
	"ref_id": "FIGREF0"
	},
	{
	"start": 572,
	"end": 580,
	"text": "Figure 5",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "The rest of this article is structured as follows: section 2 describes the theoretical and empirical aspects of document structure we model in this article. These aspects include rhetorical status and relatedness:",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "\u2022 Rhetorical status in terms of problem solving: What is the goal and contribution of the paper? This type of information is often marked by metadiscourse and by conventional patterns of presentation (cf. section 2.1).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "\u2022 Rhetorical status in terms of intellectual attribution: What information is claimed to be new, and which statements describe other work? This type of information can be recognized by following the \"agent structure\" of text, that is, by looking at all grammatical subjects occurring in sequence (cf. section 2.2).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "\u2022 Relatedness among articles: What articles is this work similar to, and in what respect? This type of information can be found by examining fixed indicator phrases like in contrast to . . . , section headers, and citations (cf. section 2.3).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "These aspects of rhetorical status are encoded in an annotation scheme that we present in section 2.4. Annotation of relevance is covered in section 2.5. In section 3, we report on the construction of a gold standard for rhetorical status and relevance and on the measurement of agreement among human annotators. We then describe in section 4 our system that simulates the human annotation. Section 5 presents an overview of the intrinsic evaluation we performed, and section 6 closes with a summary of the contribution of this work, its limitations, and suggestions for future work.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 2",
	"sec_num": null
	},
	{
	"text": "It is important for our task to find the right definition of rhetorical status to describe the content in scientific articles. The definition should both capture generalizations about the nature of scientific texts and also provide the right kind of information to enable the construction of better summaries for a practical application. Another requirement is that the analysis should be applicable to research articles from different presentational traditions and subject matters.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status, Citations, and Relevance",
	"sec_num": "2."
	},
	{
	"text": "For the development of our scheme, we used the chronologically first 80 articles in our corpus of conference articles in computational linguistics (articles presented at COLING, ANLP, and (E)ACL conferences or workshops). Because of the interdisciplinarity of the field, the papers in this collection cover a challenging range of subject matters, such as logic programming, statistical language modeling, theoretical semantics, computational dialectology, and computational psycholinguistics. Also, the research methodology and tradition of presentation is very different among these fields; (computer scientists write very different papers than theoretical linguists). We thus expect our analysis to be equally applicable in a wider range of disciplines and subdisciplines other than those named.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status, Citations, and Relevance",
	"sec_num": "2."
	},
	{
	"text": "Our model relies on the following dimensions of document structure in scientific articles.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status",
	"sec_num": "2.1"
	},
	{
	"text": "Problem structure. Research is often described as a problem-solving activity (Jordan 1984; Trawinski 1989; Zappen 1983) . Three information types can be expected to occur in any research article: problems (research goals), solutions (methods), and results. In many disciplines, particularly the experimental sciences, this problem-solution structure has been crystallized in a fixed presentation of the scientific material as introduction, method, result and discussion (van Dijk 1980) . But many texts in computational linguistics do not adhere to this presentation, and our analysis therefore has to be based on the underlying logical (rhetorical) organization, using textual representation only as an indication.",
	"cite_spans": [
	{
	"start": 77,
	"end": 90,
	"text": "(Jordan 1984;",
	"ref_id": "BIBREF38"
	},
	{
	"start": 91,
	"end": 106,
	"text": "Trawinski 1989;",
	"ref_id": "BIBREF83"
	},
	{
	"start": 107,
	"end": 119,
	"text": "Zappen 1983)",
	"ref_id": "BIBREF86"
	},
	{
	"start": 470,
	"end": 485,
	"text": "(van Dijk 1980)",
	"ref_id": "BIBREF83"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status",
	"sec_num": "2.1"
	},
	{
	"text": "Intellectual attribution. Scientific texts should make clear what the new contribution is, as opposed to previous work (specific other researchers' approaches) and background material (generally accepted statements). We noticed that intellectual attribution has a segmental character. Statements in a segment without any explicit attribution are often interpreted as belonging to the most recent explicit attribution statement (e.g., Other researchers claim that). Our rhetorical scheme assumes that readers have no difficulty in understanding intellectual attribution, an assumption that we verified experimentally.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status",
	"sec_num": "2.1"
	},
	{
	"text": "Scientific argumentation. In contrast to the view of science as a disinterested \"fact factory,\" researchers like Swales (1990) have long claimed that there is a strong social aspect to science, because the success of a researcher is correlated with her ability to convince the field of the quality of her work and the validity of her arguments. Authors construct an argument that Myers (1992) calls the \"rhetorical act of the paper\": The statement that their work is a valid contribution to science. Swales breaks down this \"rhetorical act\" into single, nonhierarchical argumentative moves (i.e., rhetorically coherent pieces of text, which perform the same communicative function). His Constructing a Research Space (CARS) model shows how patterns of these moves can be used to describe the rhetorical structure of introduction sections of physics articles. Importantly, Swales's moves describe the rhetorical status of a text segment with respect to the overall message of the document, and not with respect to adjacent text segments.",
	"cite_spans": [
	{
	"start": 113,
	"end": 126,
	"text": "Swales (1990)",
	"ref_id": "BIBREF74"
	},
	{
	"start": 380,
	"end": 392,
	"text": "Myers (1992)",
	"ref_id": "BIBREF57"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status",
	"sec_num": "2.1"
	},
	{
	"text": "Attitude toward other people's work. We are interested in how authors include reference to other work into their argument. In the flow of the argument, each piece of other work is mentioned for a specific reason: it is portrayed as a rival approach, as a prior approach with a fault, or as an approach contributing parts of the authors' own solution. In well-written papers, this relation is often expressed in an explicit way. The next section looks at the stylistic means available to the author to express the connection between previous approaches and their own work.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical Status",
	"sec_num": "2.1"
	},
	{
	"text": "Explicit metadiscourse is an integral aspect of scientific argumentation and a way of expressing attitude toward previous work. Examples for metadiscourse are phrases like we argue that and in contrast to common belief, we. Metadiscourse is ubiquitous in scientific writing: Hyland (1998) found a metadiscourse phrase on average after every 15 words in running text.",
	"cite_spans": [
	{
	"start": 275,
	"end": 288,
	"text": "Hyland (1998)",
	"ref_id": "BIBREF35"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "A large proportion of scientific metadiscourse is conventionalized, particularly in the experimental sciences, and particularly in the methodology or result section (e.g., we present original work . . . , or An ANOVA analysis revealed a marginal interaction/a main effect of . . . ). Swales (1990) lists many such fixed phrases as co-occurring with the moves of his CARS model (pages 144, 154-158, 160-161) . They are useful indicators of overall importance (Pollock and Zamora 1975) ; they can also be relatively easily recognized with information extraction techniques (e.g., regular expressions). Paice (1990) introduces grammars for pattern matching of indicator phrases, e.g., the aim/purpose of this paper/article/study and we conclude/propose.",
	"cite_spans": [
	{
	"start": 284,
	"end": 297,
	"text": "Swales (1990)",
	"ref_id": "BIBREF74"
	},
	{
	"start": 377,
	"end": 406,
	"text": "(pages 144, 154-158, 160-161)",
	"ref_id": null
	},
	{
	"start": 458,
	"end": 483,
	"text": "(Pollock and Zamora 1975)",
	"ref_id": "BIBREF61"
	},
	{
	"start": 600,
	"end": 612,
	"text": "Paice (1990)",
	"ref_id": "BIBREF59"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "Apart from this conventionalized metadiscourse, we noticed that our corpus contains a large number of metadiscourse statements that are less formalized: statements about aspects of the problem-solving process or the relation to other work. Figure 4 , for instance, shows that there are many ways to say that one's research is based on somebody else's (\"research continuation\"). The sentences do not look similar on the surface: The syntactic subject can be the authors, the originators of the method, or even the method itself. Also, the verbs are very different (base, be related, use, follow). Some sentences use metaphors of change and creation. The wide range of linguistic expression we observed presents a challenge for recognition and correct classification using standard information extraction patterns.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 240,
	"end": 248,
	"text": "Figure 4",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "With respect to agents occurring in scientific metadiscourse, we make two suggestions: (1) that scientific argumentation follows prototypical patterns and employs recurrent types of agents and actions and (2) that it is possible to recognize many of these automatically. Agents play fixed roles in the argumentation, and there are so",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 We employ Suzuki's algorithm to learn case frame patterns as dendroid distributions.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "(9605013)",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 Our method combines similarity-based estimates with Katz's back-off scheme, which is widely used for language modeling in speech recognition.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 Thus, we base our model on the work of Clark and Wilkes-Gibbs (1986), and Heeman and Hirst (1992) . . .",
	"cite_spans": [
	{
	"start": 51,
	"end": 75,
	"text": "Wilkes-Gibbs (1986), and",
	"ref_id": null
	},
	{
	"start": 76,
	"end": 86,
	"text": "Heeman and",
	"ref_id": null
	},
	{
	"start": 87,
	"end": 99,
	"text": "Hirst (1992)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 The starting point for this work was Scha and Polanyi's discourse grammar (Scha and Polanyi, 1988; Pruest et al., 1994) .",
	"cite_spans": [
	{
	"start": 76,
	"end": 100,
	"text": "(Scha and Polanyi, 1988;",
	"ref_id": null
	},
	{
	"start": 101,
	"end": 121,
	"text": "Pruest et al., 1994)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 We use the framework for the allocation and transfer of control of Whittaker and Stenton (1988) .",
	"cite_spans": [
	{
	"start": 69,
	"end": 97,
	"text": "Whittaker and Stenton (1988)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 Following Laur (1993), we consider simple prepositions (like \"in\") as well as prepositional phrases (like \"in front of\").",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "\u2022 Our lexicon is based on a finite-state transducer lexicon (Karttunen et al., 1992) . (9503004) \u2022 Instead of . . . we will adopt a simpler, monostratal representation that is more closely related to those found in dependency grammars (e.g., Hudson (1984) ).",
	"cite_spans": [
	{
	"start": 60,
	"end": 84,
	"text": "(Karttunen et al., 1992)",
	"ref_id": null
	},
	{
	"start": 235,
	"end": 255,
	"text": "(e.g., Hudson (1984)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "(9408014)",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metadiscourse and Agentivity",
	"sec_num": "2.2"
	},
	{
	"text": "Statements expressing research continuation, with source article number.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 4",
	"sec_num": null
	},
	{
	"text": "few of these roles that they can be enumerated: agents appear as rivals, as contributors of part of the solution (they), as the entire research community in the field, or as the authors of the paper themselves (we). Note the similarity of agent roles to the three kinds of intellectual attribution mentioned above. We also propose prototypical actions frequently occurring in scientific discourse: the field might agree, a particular researcher can suggest something, and a certain solution could either fail or be successful. In section 4 we will describe the three features used in our implementation that recognize metadiscourse. Another important construct that expresses relations to other researchers' work is formal citations, to which we will now turn.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 4",
	"sec_num": null
	},
	{
	"text": "Citation indexes are constructs that contain pointers between cited texts and citing texts (Garfield 1979) , traditionally in printed form. When done on-line (as in CiteSeer [Lawrence, Giles, and Bollacker 1999] , or as in Nanba and Okumura's [1999] work), citations are presented in context for users to browse. Browsing each citation is timeconsuming, but useful: just knowing that an article cites another is often not enough. One needs to read the context of the citation to understand the relation between the articles. Citations may vary in many dimensions; for example, they can be central or perfunctory, positive or negative (i.e., critical); apart from scientific reasons, there is also a host of social reasons for citing (\"politeness, tradition, piety\" [Ziman 1969]) .",
	"cite_spans": [
	{
	"start": 91,
	"end": 106,
	"text": "(Garfield 1979)",
	"ref_id": "BIBREF31"
	},
	{
	"start": 174,
	"end": 211,
	"text": "[Lawrence, Giles, and Bollacker 1999]",
	"ref_id": "BIBREF45"
	},
	{
	"start": 223,
	"end": 249,
	"text": "Nanba and Okumura's [1999]",
	"ref_id": "BIBREF58"
	},
	{
	"start": 733,
	"end": 778,
	"text": "(\"politeness, tradition, piety\" [Ziman 1969])",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Citations and Relatedness",
	"sec_num": "2.3"
	},
	{
	"text": "We concentrate on two citation contexts that are particularly important for the information needs of researchers:",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Citations and Relatedness",
	"sec_num": "2.3"
	},
	{
	"text": "\u2022 Contexts in which an article is cited negatively or contrastively.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Citations and Relatedness",
	"sec_num": "2.3"
	},
	{
	"text": "\u2022 Contexts in which an article is cited positively or in which the authors state that their own work originates from the cited work.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Citations and Relatedness",
	"sec_num": "2.3"
	},
	{
	"text": "A distinction among these contexts would enable us to build more informative citation indexes. We suggest that such a rhetorical distinction can be made manually and automatically for each citation; we use a large corpus of scientific papers along with humans' judgments of this distinction to train a system to make such distinctions.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Citations and Relatedness",
	"sec_num": "2.3"
	},
	{
	"text": "Our rhetorical annotation scheme (cf. Table 1) encodes the aspects of scientific argumentation, metadiscourse, and relatedness to other work described before. The categories are assigned to full sentences, but a similar scheme could be developed for clauses or phrases. The annotation scheme is nonoverlapping and nonhierarchical, and each sentence must be assigned to exactly one category. As adjacent sentences of the same status can be considered to form zones of the same rhetorical status, we call the units rhetorical zones. The shortest of these zones are one sentence long.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "The Rhetorical Annotation Scheme",
	"sec_num": "2.4"
	},
	{
	"text": "The rhetorical status of a sentence is determined on the basis of the global context of the paper. For instance, whereas the OTHER category describes all neutral descriptions of other researchers' work, the categories BASIS and CONTRAST are applicable to sentences expressing a research continuation relationship or a contrast to other work. Generally accepted knowledge is classified as BACKGROUND, whereas the author's own work is separated into the specific research goal (AIM) and all other statements about the author's own work (OWN). The annotation scheme expresses important discourse and argumentation aspects of scientific articles, but with its seven categories it is not designed to model the full complexity of scientific texts. The category OWN, for instance, could be further subdivided into method (solution), results, and further work, which is not done in the work reported here. There is a conflict between explanatory power and the simplicity necessary for reliable human and automatic classification, and we decided to restrict ourselves to the rhetorical distinctions that are most salient and potentially most useful for several information access applications. The user-tailored summaries and more informative citation indexes we mentioned before are just two such applications; another one is the indexing and previewing of the internal structure of the article. To make such indexing and previewing possible, our scheme contains the additional category TEXTUAL, which captures previews of section structure (section 2 describes our data . . . ). Such previews would make it possible to label sections with the author's indication of their contents.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "The Rhetorical Annotation Scheme",
	"sec_num": "2.4"
	},
	{
	"text": "Our rhetorical analysis, as noted above, is nonhierarchical, in contrast to Rhetorical Structure Theory (RST) (Mann and Thompson 1987; Marcu 1999) , and it concerns text pieces at a lower level of granularity. Although we do agree with RST that the structure of text is hierarchical in many cases, it is our belief that the relevance and function of certain text pieces can be determined without analyzing the full hierarchical structure of the text. Another difference between our analysis and that of RST is that our analysis aims at capturing the rhetorical status of a piece of text in respect to the overall message, and not in relation to adjacent pieces of text.",
	"cite_spans": [
	{
	"start": 110,
	"end": 134,
	"text": "(Mann and Thompson 1987;",
	"ref_id": "BIBREF52"
	},
	{
	"start": 135,
	"end": 146,
	"text": "Marcu 1999)",
	"ref_id": "BIBREF53"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "The Rhetorical Annotation Scheme",
	"sec_num": "2.4"
	},
	{
	"text": "As our immediate goal is to select important content from a text, we also need a second set of gold standards that are defined by relevance (as opposed to rhetorical status). Relevance is a difficult issue because it is situational to a unique occasion (Saracevic 1975; Sparck Jones 1990; Mizzaro 1997) : Humans perceive relevance differently from each other and differently in different situations. Paice and Jones (1993) report that they abandoned an informal sentence selection experiment in which they used agriculture articles and experts in the field as participants, as the participants were too strongly influenced by their personal research interest.",
	"cite_spans": [
	{
	"start": 253,
	"end": 269,
	"text": "(Saracevic 1975;",
	"ref_id": "BIBREF70"
	},
	{
	"start": 270,
	"end": 288,
	"text": "Sparck Jones 1990;",
	"ref_id": "BIBREF72"
	},
	{
	"start": 289,
	"end": 302,
	"text": "Mizzaro 1997)",
	"ref_id": "BIBREF55"
	},
	{
	"start": 400,
	"end": 422,
	"text": "Paice and Jones (1993)",
	"ref_id": "BIBREF60"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "As a result of subjectivity, a number of human sentence extraction experiments over the years have resulted in low agreement figures. Rath, Resnick, and Savage (1961) report that six participants agreed on only 8% of 20 sentences they were asked to select out of short Scientific American texts and that five agreed on 32% of the sentences. They found that after six weeks, subjects selected on average only 55% of the sentences they themselves selected previously. Edmundson and Wyllys (1961) find similarly low human agreement for research articles. More recent experiments reporting more positive results all used news text (Jing et al. 1998; Zechner 1995 ). As discussed above, the compression rates on news texts are far lower: there are fewer sentences from which to choose, making it easier to agree on which ones to select. Sentence selection from scientific texts also requires more background knowledge, thus importing an even higher level of subjectivity into sentence selection experiments.",
	"cite_spans": [
	{
	"start": 134,
	"end": 166,
	"text": "Rath, Resnick, and Savage (1961)",
	"ref_id": "BIBREF63"
	},
	{
	"start": 466,
	"end": 493,
	"text": "Edmundson and Wyllys (1961)",
	"ref_id": "BIBREF30"
	},
	{
	"start": 627,
	"end": 645,
	"text": "(Jing et al. 1998;",
	"ref_id": "BIBREF36"
	},
	{
	"start": 646,
	"end": 658,
	"text": "Zechner 1995",
	"ref_id": "BIBREF87"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Recently, researchers have been looking for more objective definitions of relevance. Kupiec, Pedersen, and Chen (1995) define relevance by abstract similarity: A sentence in a document is considered relevant if it shows a high level of similarity to a sentence in the abstract. This definition of relevance has the advantage that it is fixed (i.e., the researchers have no influence over it). It relies, however, on two assumptions: that the writing style is such that there is a high degree of overlap between sentences in the abstract and in the main text and that the abstract is indeed the target output that is most adequate for the final task.",
	"cite_spans": [
	{
	"start": 85,
	"end": 118,
	"text": "Kupiec, Pedersen, and Chen (1995)",
	"ref_id": "BIBREF42"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "In our case, neither assumption holds. First, the experiments in Teufel and Moens (1997) showed that in our corpus only 45% of the abstract sentences appear elsewhere in the body of the document (either as a close variant or in identical form), whereas Kupiec, Pedersen, and Chen report a figure of 79%. We believe that the reason for the difference is that in our case the abstracts were produced by the document authors and by professional abstractors in Kupiec, Pedersen, and Chen's case. Author summaries tend to be less systematic (Rowley 1982) and more \"deep generated,\" whereas summaries by professional abstractors follow an internalized building plan (Liddy 1991) and are often created through sentence extraction (Lancaster 1998) .",
	"cite_spans": [
	{
	"start": 65,
	"end": 88,
	"text": "Teufel and Moens (1997)",
	"ref_id": "BIBREF78"
	},
	{
	"start": 536,
	"end": 549,
	"text": "(Rowley 1982)",
	"ref_id": "BIBREF66"
	},
	{
	"start": 660,
	"end": 672,
	"text": "(Liddy 1991)",
	"ref_id": "BIBREF47"
	},
	{
	"start": 723,
	"end": 739,
	"text": "(Lancaster 1998)",
	"ref_id": "BIBREF43"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Second, and more importantly, the abstracts and improved citation indexes we intend to generate are not modeled on traditional summaries, which do not provide the type of information needed for the applications we have in mind. Information about related work plays an important role in our strategy for summarization and citation indexing, but such information is rarely found in abstracts. We empirically found that the rhetorical status of information occurring in author abstracts is very limited and consists mostly of information about the goal of the paper and specifics of the solution. Details of the analysis we conducted on this topic are given in section 3.2.2.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "We thus decided to augment our corpus with an independent set of human judgments of relevance. We wanted to replace the vague definition of relevance often used in sentence extraction experiments with a more operational definition based on rhetorical status. For instance, a sentence is considered relevant only if it describes the research goal or states a difference with a rival approach. More details of the instructions we used to make the relevance decisions are given in section 3.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Thus, we have two parallel human annotations in our corpus: rhetorical annotation and relevance selection. In both tasks, each sentence in the articles is classified: Each sentence receives one rhetorical category and also the label irrelevant or relevant. This strategy can create redundant material (e.g., when the same fact is expressed in a sentence in the introduction, a sentence in the conclusion, and one in the middle of the document). But this redundancy also helps mitigate one of the main problems with sentence-based gold standards, namely, the fact that there is no one single best extract for a document. In our annotation, all qualifying sentences in the document are identified and classified into the same group, which makes later comparisons with system performance fairer. Also, later steps cannot only find redundancy in the intermediate result and remove it, but also use the redundancy as an indication of importance.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Aim: 10 Our research addresses some of the same questions and uses similar raw data, but we investigate how to factor word association tendencies into associations of words to certain hidden senses classes and associations between the classes themselves. 22 We will consider here only the problem of classifying nouns according to their distribution as direct objects of verbs; the converse problem is formally similar. 25 The problem we study is how to use the EQN to classify the EQN. 44 In general, we are interested on how to organise a set of linguistic objects such as words according to the contexts in which they occur, for instance grammatical constructions or n-grams. 46 Our problem can be seen as that of learning a joint distribution of pairs from a large sample of pairs. 162 We have demonstrated that a general divisive clustering procedure for probability distributions can be used to group words according to their participation in particular grammatical relations with other words.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Background: 0 Methods for automatically classifying words according to their contexts of use have both scientific and practical interest. 4 The problem is that for large enough corpora the number of possible joint events is much larger than the number of event occurrences in the corpus, so many events are seen rarely or never, making their frequency counts unreliable estimates of their probabilities.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "Own ( Contrast with Other Approaches/Weaknesses of Other Approaches: 9 His notion of similarity seems to agree with our intuitions in many cases, but it is not clear how it can be used directly to construct word classes and corresponding models of association. 14 Class construction is then combinatorially very demanding and depends on frequency counts for joint events involving particular words, a potentially unreliable source of information as we noted above. 41 However, this is not very satisfactory because one of the goals of our work is precisely to avoid the problems of data sparseness by grouping words into classes.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance",
	"sec_num": "2.5"
	},
	{
	"text": "The combined entropy maximization entropy [sic] and distortion minimization is carried out by a two-stage iterative process similar to the EM method (Dempster et al., 1977) . 113 The analogy with statistical mechanics suggests a deterministic annealing procedure for clustering (Rose et al., 1990) , in which the number of clusters is determined through a sequence of phase transitions by continuously increasing the parameter EQN following an annealing schedule. 153 The data for this test was built from the training data for the previous one in the following way, based on a suggestion by Dagan et al. (1993) .",
	"cite_spans": [
	{
	"start": 149,
	"end": 172,
	"text": "(Dempster et al., 1977)",
	"ref_id": null
	},
	{
	"start": 267,
	"end": 297,
	"text": "clustering (Rose et al., 1990)",
	"ref_id": null
	},
	{
	"start": 592,
	"end": 611,
	"text": "Dagan et al. (1993)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Basis (Imported Solutions): 65",
	"sec_num": null
	},
	{
	"text": "Example of manual annotation: Relevant sentences with rhetorical status. Figure 5 gives an example of the manual annotation. Relevant sentences of all rhetorical categories are shown. Our system creates a list like the one in Figure 5 automatically ( Figure 12 shows the actual output of the system when run on the example paper). In the next section, we turn to the manual annotation step and the development of the gold standard used during system training and system evaluation.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 73,
	"end": 81,
	"text": "Figure 5",
	"ref_id": null
	},
	{
	"start": 226,
	"end": 234,
	"text": "Figure 5",
	"ref_id": null
	},
	{
	"start": 251,
	"end": 260,
	"text": "Figure 12",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Figure 5",
	"sec_num": null
	},
	{
	"text": "For any linguistic analysis that requires subjective interpretation and that is therefore not objectively true or false, it is important to show that humans share some intuitions about the analysis. This is typically done by showing that they can apply it independently of each other and that the variation they display is bounded (i.e., not arbitrarily high). The argument is strengthened if the judges are people other than the developers of the analysis, preferably \"na\u00efve\" subjects (i.e., not computational linguists). Apart from the cognitive validation of our analysis, high agreement is essential if the annotated corpus is to be used as training material for a machine learning process, like the one we describe in section 4. Noisy and unreliably annotated training material will very likely deteriorate the classification performance.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Human Judgments: The Gold Standard",
	"sec_num": "3."
	},
	{
	"text": "In inherently subjective tasks, it is also common practice to consider human performance as an upper bound. The theoretically best performance of a system is reached if agreement among a pool of human annotators does not decrease when the system is added to the pool. This is so because an automatic process cannot do any better in this situation than to be indistinguishable from human performance.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Human Judgments: The Gold Standard",
	"sec_num": "3."
	},
	{
	"text": "The annotated development corpus consists of 80 conference articles in computational linguistics (12,188 sentences; 285,934 words). It is part of a larger corpus of 260 articles (1.1 million words) that we collected from the CMP LG archive (CMP LG 1994) . The appendix lists the 80 articles (archive numbers, titles and authors) of our development corpus; it consists of the 80 chronologically oldest articles in the larger corpus, containing articles deposited between May 1994 and May 1996 (whereas the entire corpus stretches until 2001).",
	"cite_spans": [
	{
	"start": 240,
	"end": 253,
	"text": "(CMP LG 1994)",
	"ref_id": "BIBREF27"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Corpus",
	"sec_num": "3.1"
	},
	{
	"text": "Papers were included if they were presented at one of the following conferences (or associated workshops): the annual meeting of the Association for Computational Linguistics (ACL), the meeting of the European Chapter of the Association for Computational Linguistics (EACL), the conference on Applied Natural Language Processing (ANLP), the International Joint Conference on Artificial Intelligence (IJCAI), or the International Conference on Computational Linguistics (COLING). As mentioned above, a wide range of different subdomains of the field of computational linguistics are covered.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Corpus",
	"sec_num": "3.1"
	},
	{
	"text": "We added Extensible Markup Language (XML) markup to the corpus: Titles, authors, conference, date, abstract, sections, headlines, paragraphs, and sentences were marked up. Equations, tables, images were removed and replaced by placeholders. Bibliography lists were marked up and parsed. Citations and occurrences of author names in running text were recognized, and self-citations were recognized and specifically marked up. (Linguistic) example sentences and example pseudocode were manually marked up, such that clean textual material (i.e., the running text of the article without interruptions) was isolated for automatic processing. The implementation uses the Text Tokenization Toolkit (TTT) software (Grover, Mikheev, and Matheson 1999) .",
	"cite_spans": [
	{
	"start": 707,
	"end": 743,
	"text": "(Grover, Mikheev, and Matheson 1999)",
	"ref_id": "BIBREF33"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Corpus",
	"sec_num": "3.1"
	},
	{
	"text": "The annotation experiment described here (and in Teufel, Carletta, and Moens [1999] in more detail) tests the rhetorical annotation scheme presented in section 2.4.",
	"cite_spans": [
	{
	"start": 49,
	"end": 83,
	"text": "Teufel, Carletta, and Moens [1999]",
	"ref_id": "BIBREF77"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Rhetorical Status",
	"sec_num": "3.2"
	},
	{
	"text": "Annotators. Three task-trained annotators were used: Annotators A and B have degrees in cognitive science and speech therapy. They were paid for the experiment. Both are well-used to reading scientific articles for their studies and roughly understand the contents of the articles they annotated because of the closeness of their fields to computational linguistics. Annotator C is the first author. We did not want to declare annotator C the expert annotator; we believe that in subjective tasks like the one described here, there are no real experts.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "Guidelines. Written guidelines (17 pages) describe the semantics of the categories, ambiguous cases, and decision strategies. The guidelines also include the decision tree reproduced in Figure 6 .",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 186,
	"end": 194,
	"text": "Figure 6",
	"ref_id": "FIGREF2"
	}
	],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "Training. Annotators received a total of 20 hours of training. Training consisted of the presentation of annotation of six example papers and the annotation of eight training articles under real conditions (i.e., independently). In subsequent training sessions, decision criteria for difficult cases encountered in the training articles were discussed. Obviously, the training articles were excluded from measurements of human agreement.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "Materials and procedure. Twenty-five articles were used for annotation. As no annotation tool was available at the time, annotation was performed on paper; the categories were later transferred to the electronic versions of the articles by hand. Skim-reading and annotation typically took between 20 and 30 minutes per article, but there were no time restrictions. No communication between the annotators was allowed during annotation. Six weeks after the initial annotation, annotators were asked to reannotate 6 random articles out of the 25.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "Evaluation measures. We measured two formal properties of the annotation: stability and reproducibility (Krippendorff 1980) . Stability, the extent to which one annotator will produce the same classifications at different times, is important because an unstable annotation scheme can never be reproducible. Reproducibility, the extent to which different annotators will produce the same classifications, is important because it measures the consistency of shared understandings (or meaning) held between annotators. We use the kappa coefficient K (Siegel and Castellan 1988) to measure stability and reproducibility, following Carletta (1996) . The kappa coefficient is defined as follows:",
	"cite_spans": [
	{
	"start": 104,
	"end": 123,
	"text": "(Krippendorff 1980)",
	"ref_id": "BIBREF41"
	},
	{
	"start": 547,
	"end": 574,
	"text": "(Siegel and Castellan 1988)",
	"ref_id": "BIBREF71"
	},
	{
	"start": 627,
	"end": 642,
	"text": "Carletta (1996)",
	"ref_id": "BIBREF25"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "K = P(A) \u2212 P(E) 1 \u2212 P(E)",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "where P(A) is pairwise agreement and P(E) random agreement. K varies between 1 when agreement is perfect and \u22121 when there is a perfect negative correlation. K = 0 is defined as the level of agreement that would be reached by random annotation using the same distribution of categories as the actual annotators did.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "The main advantage of kappa as an annotation measure is that it factors out random agreement by numbers of categories and by their distribution. As kappa also abstracts over the number of annotators considered, it allows us to compare the agreement numerically among a group of human annotators with the agreement between the system and one or more annotators (section 5), which we use as one of the performance measures of the system.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rationale and Experimental Design.",
	"sec_num": "3.2.1"
	},
	{
	"text": "The annotation experiments show that humans distinguish the seven rhetorical categories with a stability of K = .82, .81, .76 (N = 1,220; k = 2, where K stands for the kappa coefficient, N for the number of items (sentences) annotated, and k for the number of annotators). This is equivalent to 93%, 92%, and 90% agreement. Reproducibility was measured at K = .71 (N = 4,261, k = 3), which is equivalent to 87% agreement. On Krippendorff's (1980) scale, agreement of K = .8 or above is considered as reliable, agreement of .67-.8 as marginally reliable, and less than .67 as unreliable. On Landis and Koch's (1977) more forgiving scale, agreement of .0-.2 is considered as showing \"slight\" correlation, .21-.4 as \"fair,\" .41-.6 as \"moderate,\" .61-0.8 as \"substantial,\" and .81 -1.0 as \"almost perfect.\" According to these guidelines, our results can be considered reliable, substantial annotation. Figure 7 shows that the distribution of the seven categories is very skewed, with 67% of all sentences being classified as OWN. (The distribution was calculated using all three judgments per sentence [cf. the calculation of kappa]. The total number of items is then k \u2022 N, i.e., 12,783 in this case.) Table 2 shows a confusion matrix between two annotators. The numbers represent absolute sentence numbers, and the diagonal (boldface numbers) are the counts of sentences that were identically classified by both annotators. We used Krippendorff's diagnostics to determine which particular categories humans had most problems with: For each category, agreement is measured with a new data set in which all categories except for the category of interest are collapsed into one metacategory. Original agreement is compared to that measured on the new (artificial) data set; high values show that annotators can distinguish the given category well from all others. When their results are compared to the overall reproducibility of K = .71, the annotators were good at distinguishing AIM (Krippendorff's diagnostics; K = .79) and TEXTUAL (K = .79).",
	"cite_spans": [
	{
	"start": 425,
	"end": 446,
	"text": "Krippendorff's (1980)",
	"ref_id": "BIBREF41"
	},
	{
	"start": 590,
	"end": 614,
	"text": "Landis and Koch's (1977)",
	"ref_id": "BIBREF44"
	}
	],
	"ref_spans": [
	{
	"start": 898,
	"end": 906,
	"text": "Figure 7",
	"ref_id": "FIGREF3"
	},
	{
	"start": 1199,
	"end": 1206,
	"text": "Table 2",
	"ref_id": "TABREF2"
	}
	],
	"eq_spans": [],
	"section": "Results.",
	"sec_num": "3.2.2"
	},
	{
	"text": "The high agreement in AIM sentences is a positive result that seems to be at odds with previous sentence extraction experiments. We take this as an indication that some types of rhetorical classification are easier for human minds to do than unqualified relevance decision. We also think that the positive results are partly due to the existence of the guidelines. The annotators were less consistent at determining BASIS (K = .49) and CONTRAST (K = .59). The same picture emerges if we look at precision and recall of single categories between two annotators (cf. Table 3 ). Precision and recall for AIM and TEXTUAL are high at 72%/56% and 79%/79%, whereas they are lower for CONTRAST (50%/55%) and BASIS (82%/34%).",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 565,
	"end": 572,
	"text": "Table 3",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Results.",
	"sec_num": "3.2.2"
	},
	{
	"text": "This contrast in agreement might have to do with the location of the rhetorical zones in the paper: AIM and TEXTUAL zones are usually found in fixed locations (beginning or end of the introduction section) and are explicitly marked with metadiscourse, whereas CONTRAST sentences, and even more so BASIS sentences, are usually interspersed within longer OWN zones. As a result, these categories are more exposed to lapses of attention during annotation.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Results.",
	"sec_num": "3.2.2"
	},
	{
	"text": "With respect to the longer, more neutral zones (intellectual attribution), annotators often had problems in distinguishing OTHER work from OWN work, particularly in cases where the authors did not express a clear distinction between new work and previous own work (which, according to our instructions, should be annotated as OTHER). Another persistently problematic distinction for our annotators was that between OWN Table 3 Annotator C's precision and recall per category if annotator B is gold standard.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 419,
	"end": 426,
	"text": "Table 3",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Results.",
	"sec_num": "3.2.2"
	},
	{
	"text": "Precision 72% 50% 79%",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "AIM CTR TXT OWN BKG BAS OTH",
	"sec_num": null
	},
	{
	"text": "68% 82% 74% Recall 56% 55% 79% 92% 75% 34% 83%",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "94%",
	"sec_num": null
	},
	{
	"text": "and BACKGROUND. This could be a sign that some authors aimed their papers at an expert audience and thus thought it unnecessary to signal clearly which statements are commonly agreed upon in the field, as opposed to their own new claims. If a paper is written in such a way, it can indeed be understood only with a considerable amount of domain knowledge, which our annotators did not have. Because intellectual attribution (the distinction between OWN, OTHER, and BACK-GROUND material) is an important part of our annotation scheme, we conducted a second experiment measuring how well our annotators could distinguish just these three roles, using the same annotators and 22 different articles. We wrote seven pages of new guidelines describing the semantics of the three categories. Results show higher stability compared to the full annotation scheme (K = .83, .79, .81 ; N = 1,248; k = 2) and higher reproducibility (K = .78, N = 4,031, k = 3), corresponding to 94%, 93%, and 93% agreement (stability) and 93% (reproducibility). It is most remarkable that agreement of annotation of intellectual attribution in the abstracts is almost perfect: K = .98 (N = 89, k = 3), corresponding to 99% agreement. This points to the fact that authors, when writing abstracts for their papers, take care to make it clear to whom a certain statement is attributed. This effect also holds for the annotation with the full scheme with all seven categories: again, reproducibility in the abstract is higher (K = .79) than in the entire document (K = .71), but the effect is much weaker.",
	"cite_spans": [
	{
	"start": 854,
	"end": 872,
	"text": "(K = .83, .79, .81",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "94%",
	"sec_num": null
	},
	{
	"text": "Abstracts might be easier to annotate than the rest of a paper, but this does not necessarily make it possible to define a gold standard solely by looking at the abstracts. As foreshadowed in section 2.5, abstracts do not contain all types of rhetorical information. AIM and OWN sentences make up 74% of the sentences in abstracts, and only 5% of all CONTRAST sentences and 3% of all BASIS sentences occur in the abstract.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "94%",
	"sec_num": null
	},
	{
	"text": "Abstracts in our corpus are also not structurally homogeneous. When we inspected the rhetorical structure of abstracts in terms of sequences of rhetorical zones, we found a high level of variation. Even though the sequence AIM-OWN is very common (contained in 73% of all abstracts), the 80 abstracts still contain 40 different rhetorical sequences, 28 of which are unique. This heterogeneity is in stark contrast to the systematic structures Liddy (1991) found to be produced by professional abstractors. Both observations, the lack of certain rhetorical types in the abstracts and their rhetorical heterogeneity, reassure us in our decision not to use human-written abstracts as a gold standard.",
	"cite_spans": [
	{
	"start": 442,
	"end": 454,
	"text": "Liddy (1991)",
	"ref_id": "BIBREF47"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "94%",
	"sec_num": null
	},
	{
	"text": "We collected two different kinds of relevance gold standards for the documents in our development corpus: abstract-similar document sentences and additional manually selected sentences.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "In order to establish alignment between summary and document sentences, we used a semiautomatic method that relies on a simple surface similarity measure (longest common subsequence of content words, i.e., excluding words on a stop list). As in Kupiec, Pedersen, and Chen's experiment, final alignment was decided by a human judge, and the criterion was semantic similarity of the two sentences. The following sentence pair illustrates a direct match:",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "Summary: In understanding a reference, an agent determines his confidence in its adequacy as a means of identifying the referent. Document: An agent understands a reference once he is confident in the adequacy of its (inferred) plan as a means of identifying the referent.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "Of the 346 abstract sentences contained in the 80 documents, 156 (45%) could be aligned this way. Because of this low agreement and because certain rhetorical types are not present in the abstracts, we decided not to rely on abstract alignment as our only gold standard. Instead, we used manually selected sentences as an alternative gold standard, which is more informative, but also more subjective.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "We wrote eight pages of guidelines that describe relevance criteria (e.g., our definition prescribes that neutral descriptions of other work be selected only if the other work is an essential part of the solution presented, whereas all statements of criticism are to be included). The first author annotated all documents in the development corpus for relevance using the rhetorical zones and abstract similarity as aides in the relevance decision, and also skim-reading the whole paper before making the decision. This resulted in 5 to 28 sentences per paper and a total of 1,183 sentences.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "Implicitly, rhetorical classification of the extracted sentences was already given as each of these sentences already had a rhetorical status assigned to it. However, the rhetorical scheme we used for this task is slightly different. We excluded TEXTUAL, as this category was designed for document uses other than summarization. If a selected sentence had the rhetorical class TEXTUAL, it was reclassified into one of the other six categories. Figure 8 shows the resulting category distribution among these 1,183 sentences, which is far more evenly distributed than the one covering all sentences (cf. Figure 7 ). CONTRAST and OWN are the two most frequent categories.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 444,
	"end": 452,
	"text": "Figure 8",
	"ref_id": null
	},
	{
	"start": 602,
	"end": 610,
	"text": "Figure 7",
	"ref_id": "FIGREF3"
	}
	],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "We did not verify the relevance annotation with human experiments. We accept that the set of sentences chosen by the human annotator is only one possible gold standard. What is more important is that humans can agree on the rhetorical status of the relevant sentences. Liddy observed that agreement on rhetorical status was easier for professional abstractors than sentence selection: Although they did not necessarily agree on which individual sentences should go into an abstract, they did agree on the rhetorical information types that make up a good abstract.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "We asked our trained annotators to classify a set of 200 sentences, randomly sampled from the 1,183 sentences selected by the first author, into the six rhetorical categories. The sentences were presented in order of occurrence in the document, but without any context in terms of surrounding sentences. We measured stability at K = .9, .86, .83 (N = 100, k = 2) and reproducibility at K = .84 (N = 200, k = 3). These results are reassuring: They show that the rhetorical status for important sentences can be particularly well determined, better than rhetorical status for all sentences in the document (for which reproducibility was K = .71; cf. section 3.2.2).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation of Relevance",
	"sec_num": "3.3"
	},
	{
	"text": "Distribution of rhetorical categories (relevant sentences).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 8",
	"sec_num": null
	},
	{
	"text": "We now describe an automatic system that can perform extraction and classification of rhetorical status on unseen text (cf. also a prior version of the system reported in Teufel and Moens [2000] and Teufel [1999] ). We decided to use machine learning to perform this extraction and classification, based on a variety of sentential features similar to the ones reported in the sentence extraction literature. Human annotation is used as training material such that the associations between these sentential features and the target sentences can be learned. It is also used as gold standard for intrinsic system evaluation.",
	"cite_spans": [
	{
	"start": 171,
	"end": 194,
	"text": "Teufel and Moens [2000]",
	"ref_id": "BIBREF79"
	},
	{
	"start": 199,
	"end": 212,
	"text": "Teufel [1999]",
	"ref_id": "BIBREF75"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "The System",
	"sec_num": "4."
	},
	{
	"text": "A simpler machine learning approach using only word frequency information and no other features, as typically used in tasks like text classification, could have been employed (and indeed Nanba and Okumura [1999] do so for classifying citation contexts). To test if such a simple approach would be enough, we performed a text categorization experiment, using the Rainbow implementation of a na\u00efve Bayes term frequency times inverse document frequency (TF*IDF) method (McCallum 1997) and considering each sentence as a \"document.\" The result was a classification performance of K = .30; the classifier nearly always chooses OWN and OTHER segments. The rare but important categories AIM, BACKGROUND, CONTRAST, and BASIS could be retrieved only with low precision and recall. Therefore, text classification methods do not provide a solution to our problem. This is not surprising, given that the definition of our task has little to do with the distribution of \"content-bearing\" words and phrases, much less so than the related task of topic segmentation (Morris and Hirst 1991; Hearst 1997; Choi 2000) , or Saggion and Lapalme's (2000) approach to the summarization of scientific articles, which relies on scientific concepts and their relations. Instead, we predict that other indicators apart from the simple words contained in the sentence could provide strong evidence for the modeling of rhetorical status. Also, the relatively small amount of training material we have at our disposal requires a machine learning method that makes optimal use of as many different kinds of features as possible. We predicted that this would increase precision and recall on the categories in which we are interested. The text classification experiment is still useful as it provides a nontrivial baseline for comparison with our intrinsic system evaluation presented in section 5.",
	"cite_spans": [
	{
	"start": 187,
	"end": 211,
	"text": "Nanba and Okumura [1999]",
	"ref_id": "BIBREF58"
	},
	{
	"start": 466,
	"end": 481,
	"text": "(McCallum 1997)",
	"ref_id": "BIBREF54"
	},
	{
	"start": 1051,
	"end": 1074,
	"text": "(Morris and Hirst 1991;",
	"ref_id": "BIBREF56"
	},
	{
	"start": 1075,
	"end": 1087,
	"text": "Hearst 1997;",
	"ref_id": "BIBREF34"
	},
	{
	"start": 1088,
	"end": 1098,
	"text": "Choi 2000)",
	"ref_id": "BIBREF26"
	},
	{
	"start": 1104,
	"end": 1132,
	"text": "Saggion and Lapalme's (2000)",
	"ref_id": "BIBREF67"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "The System",
	"sec_num": "4."
	},
	{
	"text": "We use a na\u00efve Bayesian model as in Kupiec, Pedersen, and Chen's (1995) experiment (cf. Figure 9 ). Sentential features are collected for each sentence (Table 4 gives an overview of the features we used). Learning is supervised: In the training phase, associations between these features and human-provided target categories are learned. The target categories are the seven categories in the rhetorical annotation experiment and relevant/nonrelevant in the relevance selection experiment. In the testing phase, the trained model provides the probability of each target category for each sentence of unseen text, on the basis of the sentential features identified for the sentence.",
	"cite_spans": [
	{
	"start": 36,
	"end": 71,
	"text": "Kupiec, Pedersen, and Chen's (1995)",
	"ref_id": "BIBREF42"
	}
	],
	"ref_spans": [
	{
	"start": 88,
	"end": 96,
	"text": "Figure 9",
	"ref_id": null
	},
	{
	"start": 152,
	"end": 160,
	"text": "(Table 4",
	"ref_id": "TABREF3"
	}
	],
	"eq_spans": [],
	"section": "Classifiers",
	"sec_num": "4.1"
	},
	{
	"text": "Some of the features in our feature pool are unique to our approach, for instance, the metadiscourse features. Others are borrowed from the text extraction literature (Paice 1990) or related tasks and adapted to the problem of determining rhetorical status.",
	"cite_spans": [
	{
	"start": 167,
	"end": 179,
	"text": "(Paice 1990)",
	"ref_id": "BIBREF59"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "Absolute location of a sentence. In the news domain, sentence location is the single most important feature for sentence selection (Brandow, Mitze, and Rau 1995); in our domain, location information, although less dominant, can still give a useful indication. Rhetorical zones appear in typical positions in the article, as scientific argumentation",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "P(C \| F 0 , . . . , F n\u22121 ) \u2248 P(C) n\u22121 j=0 P(F j \| C) n\u22121 j=0 P(F j ) P(C \| F 0 , . . . , F n\u22121 ):",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "Probability that a sentence has target category C, given its feature values F 0 , . . . , F n\u22121 ; P(C):",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "(Overall) probability of category C; P(F j \| C):",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "Probability of feature-value pair F j , given that the sentence is of target category C; P(F j ):",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "Probability of feature value F j ;",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Features",
	"sec_num": "4.2"
	},
	{
	"text": "Na\u00efve Bayesian classifier. follows certain patterns (Swales 1990 ). For example, limitations of the author's own method can be expected to be found toward the end of the article, whereas limitations of other researchers' work are often discussed in the introduction. We observed that the size of rhetorical zones depends on location, with smaller rhetorical zones occurring toward the beginning and the end of the article. We model this by assigning location values in the following fashion: The article is divided into 20 equal parts, counting sentences. Sentences occurring in parts 1, 2, 3, 4, 19, and 20 receive the values A, B, C, D, I, and J, respectively. Parts 5 and 6 are pooled, and sentences occurring in them are given the value E; the same procedure is applied to parts 15 and 16 (value G) and 17 and 18 (value H). The remaining sentences in the middle (parts 7-14) all receive the value F (cf. Figure 10) . Section structure. Sections can have an internal structuring; for instance, sentences toward the beginning of a section often have a summarizing function. The section location feature divides each section into three parts and assigns seven values: first sentence, last sentence, second or third sentence, second-last or third-last sentence, or else either somewhere in the first, second, or last third of the section.",
	"cite_spans": [
	{
	"start": 52,
	"end": 64,
	"text": "(Swales 1990",
	"ref_id": "BIBREF74"
	}
	],
	"ref_spans": [
	{
	"start": 908,
	"end": 918,
	"text": "Figure 10)",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "Paragraph structure. In many genres, paragraphs also have internal structure (Wiebe 1994), with high-level or summarizing sentences occurring more often at the periphery of paragraphs. In this feature, sentences are distinguished into those leading or ending a paragraph and all others.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "Headlines. Prototypical headlines can be an important predictor of the rhetorical status of sentences occurring in the given section; however, not all texts in our collection use such headlines. Whenever a prototypical headline is recognized (using a set of regular expressions), it is classified into one of the following 15 classes: Introduction, Implementation, Example, Conclusion, Result, Evaluation, Solution, Experiment, Discussion, Method, Problems, Related Work, Data, Further Work, Problem Statement. If none of the patterns match, the value Non-Prototypical is assigned.",
	"cite_spans": [
	{
	"start": 365,
	"end": 510,
	"text": "Example, Conclusion, Result, Evaluation, Solution, Experiment, Discussion, Method, Problems, Related Work, Data, Further Work, Problem Statement.",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "Sentence length. Kupiec, Pedersen, and Chen (1995) report sentence length as a useful feature for text extraction. In our implementation, sentences are divided into long or short sentences, by comparison to a fixed threshold (12 words).",
	"cite_spans": [
	{
	"start": 17,
	"end": 50,
	"text": "Kupiec, Pedersen, and Chen (1995)",
	"ref_id": "BIBREF42"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "Title word contents. Sentences containing many \"content-bearing\" words have been hypothesized to be good candidates for text extraction. Baxendale (1958) extracted all words except those on the stop list from the title and the headlines and determined for each sentence whether or not it contained these words. We received better results by excluding headline words and using only title words.",
	"cite_spans": [
	{
	"start": 137,
	"end": 153,
	"text": "Baxendale (1958)",
	"ref_id": "BIBREF23"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "TFIDF word contents. How content-bearing a word is can also be measured with frequency counts (Salton and McGill 1983) . The TFIDF formula assigns high values to words that occur frequently in one document, but rarely in the overall collection of documents. We use the 18 highest-scoring TF*IDF words and classify sentences into those that contain one or more of these words and those that do not. Verb syntax. Linguistic features like tense and voice often correlate with rhetorical zones; Biber (1995) and Riley (1991) show correlation of tense and voice with prototypical section structure (\"method,\" \"introduction\"). In addition, the presence or absence ",
	"cite_spans": [
	{
	"start": 95,
	"end": 119,
	"text": "(Salton and McGill 1983)",
	"ref_id": "BIBREF68"
	},
	{
	"start": 493,
	"end": 505,
	"text": "Biber (1995)",
	"ref_id": "BIBREF24"
	},
	{
	"start": 510,
	"end": 522,
	"text": "Riley (1991)",
	"ref_id": "BIBREF65"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 9",
	"sec_num": null
	},
	{
	"text": "Values of location feature. of a modal auxiliary might be relevant for detecting the phenomenon of \"hedging\" (i.e., statements in which an author distances herself from her claims or signals low certainty: these results might indicate that . . . possibly . . . [Hyland 1998 ]). For each sentence, we use part-of-speech-based heuristics to determine tense, voice, and presence of modal auxiliaries. This algorithm is shared with the metadiscourse features, and the details are described below.",
	"cite_spans": [
	{
	"start": 261,
	"end": 273,
	"text": "[Hyland 1998",
	"ref_id": "BIBREF35"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Citation. There are many connections between citation behavior and relevance or rhetorical status. First, if a sentence contains a formal citation or the name of another author mentioned in the bibliography, it is far more likely to talk about other work than about own work. Second, if it contains a self-citation, it is far more likely to contain a direct statement of continuation (25%) than a criticism (3%). Third, the importance of a citation has been related to the distinction between authorial and parenthetical citations. Citations are called authorial if they form a syntactically integral part of the sentence or parenthetical if they do not (Swales 1990 ). In most cases, authorial citations are used as the subject of a sentence, and parenthetical ones appear toward the middle or the end of the sentence.",
	"cite_spans": [
	{
	"start": 654,
	"end": 666,
	"text": "(Swales 1990",
	"ref_id": "BIBREF74"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "We built a recognizer for formal citations. It parses the reference list at the end of the article and determines whether a citation is a self-citation (i.e., if there is an overlap between the names of the cited researchers and the authors of the current paper), and it also finds occurrences of authors' names in running text, but outside of formal citation contexts (e.g., Chomsky also claims that . . . ). The citation feature reports whether a sentence contains an author name, a citation, or nothing. If it contains a citation, the value records whether it is a self-citation and also records the location of the citation in the sentence (in the beginning, the middle, or the end). This last distinction is a heuristic for the authorial/parenthetical distinction. We also experimented with including the number of different citations in a sentence, but this did not improve results.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "History. As there are typical patterns in the rhetorical zones (e.g., AIM sentences tend to follow CONTRAST sentences), we wanted to include the category assigned to the previous sentence as one of the features. In unseen text, however, the previous target is unknown at training time (it is determined during testing). It can, however, be calculated as a second pass process during training. In order to avoid a full Viterbi search of all possibilities, we perform a beam search with width of three among the candidates of the previous sentence, following Barzilay et al. (2000) .",
	"cite_spans": [
	{
	"start": 557,
	"end": 579,
	"text": "Barzilay et al. (2000)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Formulaic expressions. We now turn to the last three features in our feature pool, the metadiscourse features, which are more sophisticated than the other features. The first metadiscourse feature models formulaic expressions like the ones described by Swales, as they are semantic indicators that we expect to be helpful for rhetorical classification. We use a list of phrases described by regular expressions, similar to Paice's (1990) grammar. Our list is divided into 18 semantic classes (cf. Table 5 ), comprising a total of 644 patterns. The fact that phrases are clustered is a simple way of dealing with data sparseness. In fact, our experiments in section 5.1.2 will show the usefulness of our (manual) semantic clusters: The clustered list performs much better than the unclustered list (i.e., when the string itself is used as a value instead of its semantic class).",
	"cite_spans": [
	{
	"start": 423,
	"end": 437,
	"text": "Paice's (1990)",
	"ref_id": "BIBREF59"
	}
	],
	"ref_spans": [
	{
	"start": 497,
	"end": 504,
	"text": "Table 5",
	"ref_id": "TABREF5"
	}
	],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Agent. Agents and actions are more challenging to recognize. We use a mechanism that, dependent on the voice of a sentence, recognizes agents (subjects or prepositional phrases headed by by) and their predicates (\"actions\"). Classification of agents and actions relies on a manually created lexicon of manual classes. As in the Formulaic feature, similar agents and actions are generalized and clustered together to avoid data sparseness. The lexicon for agent patterns (cf. Table 6 ) contains 13 types of agents and a total of 167 patterns. These 167 patterns expand to many more strings as we use a replace mechanism (e.g., the placeholder WORK NOUN in the sixth row of Table 6 can be replaced by a set of 37 nouns including theory, method, prototype, algorithm).",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 475,
	"end": 482,
	"text": "Table 6",
	"ref_id": "TABREF6"
	},
	{
	"start": 672,
	"end": 679,
	"text": "Table 6",
	"ref_id": "TABREF6"
	}
	],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "The main three agent types we distinguish are US AGENT, THEM AGENT, and GEN-ERAL AGENT, following the types of intellectual attribution discussed above. A fourth type is US PREVIOUS AGENT (the authors, but in a previous paper).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Additional agent types include nonpersonal agents like aims, problems, solutions, absence of solution, or textual segments. There are four equivalence classes of agents with ambiguous reference (\"this system\"): REF AGENT, REF US AGENT, THEM PRO-NOUN AGENT, and AIM REF AGENT. Agent classes were created based on intuition, but subsequently each class was tested with corpus statistics to determine whether it should be removed or not. We wanted to find and exclude classes that had a distribution very similar to the overall distribution of the target categories, as such features are not distinctive. We measured associations using the log-likelihood measure (Dunning 1993) for each combination of target category and semantic class by converting each cell of the contingency into a 2\u00d72 contingency table. We kept only classes of verbs in which at least one category showed a high association (gscore > 5.0), as that means that in these cases the distribution was significantly different from the overall distribution. The last column in Table 6 shows that the classes THEM PRONOUN, GENERAL, SOLUTION, PROBLEM, and REF were removed; removal improved the performance of the Agent feature.",
	"cite_spans": [
	{
	"start": 660,
	"end": 674,
	"text": "(Dunning 1993)",
	"ref_id": "BIBREF28"
	}
	],
	"ref_spans": [
	{
	"start": 211,
	"end": 275,
	"text": "REF AGENT, REF US AGENT, THEM PRO-NOUN AGENT, and AIM REF AGENT.",
	"ref_id": null
	},
	{
	"start": 1039,
	"end": 1046,
	"text": "Table 6",
	"ref_id": "TABREF6"
	}
	],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "SegAgent. SegAgent is a variant of the Agent feature that keeps track of previously recognized agents; unmarked sentences receive these previous agents as a value (in the Agent feature, they would have received the value None).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Action. We use a manually created action lexicon containing 365 verbs (cf. Table 7 ). The verbs are clustered into 20 classes based on semantic concepts such as similarity, contrast, competition, presentation, argumentation, and textual structure. For example, PRESENTATION ACTIONs include communication verbs like present, report, and state (Myers 1992; Thompson and Yiyun 1991) , RESEARCH ACTIONS include analyze, conduct, define and observe, and ARGUMENTATION ACTIONS include argue, disagree, and object to. Domain-specific actions are contained in the classes indicating a problem (fail, degrade, waste, overestimate) and solution-contributing actions (circumvent, solve, mitigate). The recognition of negation is essential; the semantics of not solving is closer to being problematic than it is to solving.",
	"cite_spans": [
	{
	"start": 342,
	"end": 354,
	"text": "(Myers 1992;",
	"ref_id": "BIBREF57"
	},
	{
	"start": 355,
	"end": 379,
	"text": "Thompson and Yiyun 1991)",
	"ref_id": "BIBREF80"
	}
	],
	"ref_spans": [
	{
	"start": 75,
	"end": 82,
	"text": "Table 7",
	"ref_id": "TABREF7"
	}
	],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "The following classes were removed by the gscore test described above, because their distribution was too similar to the overall distribution: FUTURE INTEREST, NEED, ARGUMENTATION, AFFECT in both negative and positive contexts (X in last column of Table 7) , and AWARENESS only in positive context (+ in last column). The following classes had too few occurrences in negative context (< 10 occurrences in the whole verb class) and thus the negative context of the class was also removed: BETTER SOLUTION, CONTRAST, PRESENTATION, PROBLEM (\u2212 in last column). Again, the removal improved the performance of the Action feature.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 248,
	"end": 256,
	"text": "Table 7)",
	"ref_id": "TABREF7"
	}
	],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "The algorithm for determining agents and actions relies on finite-state patterns over part-of-speech (POS) tags. Starting from each finite verb, the algorithm collects chains of auxiliaries belonging to the associated finite clause and thus determines the clause's tense and voice. Other finite verbs and commas are assumed to be clause boundaries. Once the semantic verb is found, its stem is looked up in the action lexicon. Negation is determined if one of 32 fixed negation words is present in a six-word window to the right of the finite verb.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "As our classifier requires one unique value for each classified item for each feature, we had to choose one value for sentences containing more than one finite clause. We return the following values for the action and agents feature: the first agent/action pair, if both are nonzero, otherwise the first agent without an action, otherwise the first action without an agent, if available.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "In order to determine the level of correctness of agent and action recognition, we had first to evaluate manually the error level of the POS tagging of finite verbs, as our algorithm crucially relies on finite verbs. In a random sample of 100 sentences from our development corpus that contain any finite verbs at all (they happened to contain a total of 184 finite verbs), the tagger (which is part of the TTT software) showed a recall of 95% and a precision of 93%.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "We found that for the 174 correctly determined finite verbs, the heuristics for negation and presence of modal auxiliaries worked without any errors (100% accuracy, eight negated sentences). The correct semantic verb was determined with 96% accuracy; most errors were due to misrecognition of clause boundaries. Action Type lookup was fully correct (100% accuracy), even in the case of phrasal verbs and longer idiomatic expressions (have to is a NEED ACTION; be inspired by is a CONTINUE ACTION). There were seven voice errors, two of which were due to POS-tagging errors (past participle misrecognized). The remaining five voice errors correspond to 98% accuracy.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Correctness of Agent Type determination was tested on a random sample of 100 sentences containing at least one agent, resulting in 111 agents. No agent pattern that should have been identified was missed (100% recall). Of the 111 agents, 105 cases were correct (precision of 95%). Therefore, we consider the two features to be adequately robust to serve as sentential features in our system.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Having detailed the features and classifiers of the machine learning system we use, we will now turn to an intrinsic evaluation of its performance.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 10",
	"sec_num": null
	},
	{
	"text": "Our task is to perform content selection from scientific articles, which we do by classifying sentences into seven rhetorical categories. The summaries based on this classification use some of these sentences directly, namely, sentences that express the contribution of a particular article (AIM), sentences expressing contrasts with other work (CONTRAST), and sentences stating imported solutions from other work (BASIS). Other, more frequent rhetorical categories, namely OTHER, OWN, and BACKGROUND, might also be extracted into the summary.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Intrinsic System Evaluation",
	"sec_num": "5."
	},
	{
	"text": "Because the task is a mixture of extraction and classification, we report system performance as follows:",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Intrinsic System Evaluation",
	"sec_num": "5."
	},
	{
	"text": "\u2022 We first report precision and recall values for all categories, in comparison to human performance and the text categorization baseline, as we are primarily interested in good performance on the categories AIM, CONTRAST, BASIS, and BACKGROUND.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Intrinsic System Evaluation",
	"sec_num": "5."
	},
	{
	"text": "\u2022 We are also interested in good overall classification performance, which we report using kappa and macro-F as our metric. We also discuss how well each single features does in the classification.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Intrinsic System Evaluation",
	"sec_num": "5."
	},
	{
	"text": "\u2022 We then compare the extracted sentences to our human gold standard for relevance and report the agreement in precision and agreement per category.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Intrinsic System Evaluation",
	"sec_num": "5."
	},
	{
	"text": "The results of stochastic classification were compiled with a 10-fold cross-validation on our 80-paper corpus. As we do not have much annotated material, cross-validation is a practical way to test as it can make use of the full development corpus for training, without ever using the same data for training and testing. Figure 11 show that the stochastic model obtains substantial improvement over the baseline in terms of precision and recall of the important categories AIM, BACKGROUND, CONTRAST, and BASIS. We use the F-measure, defined by van Rijsbergen (1979) as 2PR P+R , as a convenient way of reporting precision (P) and recall (R) in one value. F-measures for our categories range from .61 (TEXTUAL) and .52 (AIM) to .45 (BACKGROUND), .38 (BASIS), and .26 (CONTRAST). The recall for some categories is relatively low. As our gold standard is designed to contain a lot of redundant information for the same category, this is not too worrying. Low precision in some categories (e.g., 34% for CONTRAST, in contrast to human precision of 50%), however, could potentially present a problem for later steps in the document summarization process.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 321,
	"end": 330,
	"text": "Figure 11",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Determination of Rhetorical Status",
	"sec_num": "5.1"
	},
	{
	"text": "Overall, we find these results encouraging, particularly in view of the subjective nature of the task and the high compression achieved (2% for AIM, BASIS, and TEXTUAL sentences, 5% for CONTRAST sentences, and 6% for BACKGROUND sentences). No direct comparison with Kupiec, Pedersen, and Chen's results is possible as different data sets are used and as Kupiec et al.'s relevant sentences do not directly map into one of our categories. Assuming, however, that their relevant sentences are probably most ",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Overall Results. Table 8 and",
	"sec_num": "5.1.1"
	},
	{
	"text": "Performance per category: F-measure. comparable to our AIM sentences, our precision and recall of 44% and 65% compare favorably to theirs (42% and 42%). Table 9 shows a confusion matrix between one annotator and the system. The system is likely to confuse AIM and OWN sentences (e.g., 100 out of 172 sentences incorrectly classified as AIM by the system turned out to be OWN sentences). It also shows a tendency to confuse OTHER and OWN sentences. The system also fails to distinguish categories involving other people's work (e.g. OTHER, BASIS, and CONTRAST). Overall, these tendencies mirror human errors, as can be seen from a comparison with Table 2 . Table 10 shows the results in terms of three overall measures: kappa, percentage accuracy, and macro-F (following Lewis [1991] ). Macro-F is the mean of the F-measures of all seven categories. One reason for using macro-F and kappa is that we want to measure success particularly on the rare categories that are needed for our final task (i.e., AIM, BASIS, and CONTRAST). Microaveraging techniques like traditional accuracy tend to overestimate the contribution of frequent categories in skewed distributions like ours; this is undesirable, as OWN is the least interesting category for our purposes. This situation has parallels in information retrieval, where precision and recall are used because accuracy overestimates the performance on irrelevant items. In the case of macro-F, each category is treated as one unit, independent of the number of items contained in it. Therefore, the classification success of the individual items in rare categories is given more importance than the classification success of frequent-category items. When looking at the numerical values, however, one should keep in mind that macroaveraging results are in general numerically lower (Yang and Liu 1999) . This is because there are fewer training cases for the rare categories, which therefore perform worse with most classifiers.",
	"cite_spans": [
	{
	"start": 776,
	"end": 782,
	"text": "[1991]",
	"ref_id": null
	},
	{
	"start": 1827,
	"end": 1846,
	"text": "(Yang and Liu 1999)",
	"ref_id": "BIBREF85"
	}
	],
	"ref_spans": [
	{
	"start": 153,
	"end": 160,
	"text": "Table 9",
	"ref_id": "TABREF9"
	},
	{
	"start": 646,
	"end": 653,
	"text": "Table 2",
	"ref_id": "TABREF2"
	},
	{
	"start": 656,
	"end": 664,
	"text": "Table 10",
	"ref_id": "TABREF0"
	}
	],
	"eq_spans": [],
	"section": "Figure 11",
	"sec_num": null
	},
	{
	"text": "In the case of kappa, classifications that incorrectly favor frequent categories are punished because of a high random agreement. This effect can be shown most easily when the baselines are considered. The most ambitious baseline we use is the output of a text categorization system, as described in section 4. Other possible baselines, which are all easier to beat, include classification by the most frequent category. This baseline turns out to be trivial, as it does not extract any of the rare rhetorical categories in which we are particularly interested, and therefore receives a low kappa value at K = \u2212.12. Possible chance baselines include random annotation with uniform distribution (K = \u2212.10; accuracy of 14%) and random annotation with observed distribution. The latter baseline is built into the definition of kappa (K = 0; accuracy of 48%).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 11",
	"sec_num": null
	},
	{
	"text": "Although our system outperforms an ambitious baseline (macro-F shows that our system performs roughly 20% better than text classification) and also performs much above chance, there is still a big gap in performance between humans and machine. Macro-F shows a 20% difference between our system and human performance. If the system is put into a pool of annotators for the 25 articles for which three-way human judgment exists, agreement drops from K = .71 to K = .59. This is a clear indication that the system's annotation is still distinguishably different from human annotation.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 11",
	"sec_num": null
	},
	{
	"text": "The previous results were compiled using all features, which is the optimal feature combination (as determined by an exhaustive search in the space of feature combinations). The most distinctive single feature is Location (achieving an agreement of K = .22 against one annotator, if this feature is used as the sole feature), followed by SegAgent (K = .19), Citations (K = .18), Headlines (K = .17), Agent (K = .08), and Formulaic (K = .07). In each case, the unclustered versions of Agent, SegAgent, and Formulaic performed much worse than the clustered versions; they did not improve final results when added into the feature pool.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Feature Impact.",
	"sec_num": "5.1.2"
	},
	{
	"text": "Action performs slightly better at K = \u2212.11 than the baseline by most frequent category, but far worse than random by observed distribution. The following features on their own classify each sentence as OWN (and therefore achieve K = \u2212.12): Relative Location, Paragraphs, TF*IDF, Title, Sentence Length, Modality, Tense, and Voice. History performs very badly on its own at K = \u2212.51; it classifies almost all sentences as BACKGROUND. It does this because the probability of the first sentence's being a BACKGROUND sentence is almost one, and, if no other information is available, it is very likely that another BACKGROUND sentence will follow after a BACKGROUND sentence.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Feature Impact.",
	"sec_num": "5.1.2"
	},
	{
	"text": "Each of these features, however, still contributes to the final result: If any of them is taken out of the feature pool, classification performance decreases. How can this be, given that the individual features perform worse than chance? As the classifier derives the posterior probability by multiplying evidence from each feature, even slight evidence coming from one feature can direct the decision in the right direction. A feature that contributes little evidence on its own (too little to break the prior probability, which is strongly biased toward OWN) can thus, in combination with others, still help in disambiguating. For the na\u00efve Bayesian classification method, indeed, it is most important that the features be as independent of each other as possible. This property cannot be assessed by looking at the feature's isolated performance, but only in combination with others.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Feature Impact.",
	"sec_num": "5.1.2"
	},
	{
	"text": "It is also interesting to see that certain categories are disambiguated particularly well by certain features (cf. Table 11 ). The Formulaic feature, which is by no means the strongest feature, is nevertheless the most diverse, as it contributes to the disambiguation of six categories directly. This is because many different rhetorical categories have typical cue phrases associated with them (whereas not all categories have a preferred location in the document). Not surprisingly, Location and History are the features particularly useful for detecting BACKGROUND sentences, and SegAgent additionally contributes toward the determination of BACKGROUND zones (along with the Formulaic and the Absolute Location features). The Agent and Action features also prove their worth as they manage to disambiguate categories that many of the other features alone cannot disambiguate (e.g., CONTRAST).",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 115,
	"end": 123,
	"text": "Table 11",
	"ref_id": "TABREF0"
	}
	],
	"eq_spans": [],
	"section": "Feature Impact.",
	"sec_num": "5.1.2"
	},
	{
	"text": "The Example Paper. In order to give the reader an impression of how the figures reported in the previous section translate into real output, we present in figure 12 the output of the system when run on the example paper (all AIM, CONTRAST, and BASIS sentences). The second column shows whether the human judge agrees with the system's decision (a tick for correct decisions, and the human's preferred category for incorrect decisions). Ten out of the 15 extracted sentences have been classified correctly.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "System Output:",
	"sec_num": "5.1.3"
	},
	{
	"text": "The example also shows that the determination of rhetorical status is not always straightforward. For example, whereas the first AIM sentence that the system proposes (sentence 8) is clearly wrong, all other \"incorrect\" AIM sentences carry important in- (Resnik, 1992) Hindle's parser Fidditch (Hindle, 1993) . \u221a 20 More recently, we have constructed similar tables with the help of a statistical part-of-speech tagger (Church, 1988) and of tools for regular expression pattern matching on tagged corpora (Yarowsky, 1992) . \u221a * 113 The analogy with statistical mechanics suggests a deterministic annealing procedure for clustering (Rose et al., 1990) ",
	"cite_spans": [
	{
	"start": 254,
	"end": 268,
	"text": "(Resnik, 1992)",
	"ref_id": null
	},
	{
	"start": 269,
	"end": 308,
	"text": "Hindle's parser Fidditch (Hindle, 1993)",
	"ref_id": null
	},
	{
	"start": 419,
	"end": 433,
	"text": "(Church, 1988)",
	"ref_id": null
	},
	{
	"start": 505,
	"end": 521,
	"text": "(Yarowsky, 1992)",
	"ref_id": null
	},
	{
	"start": 631,
	"end": 650,
	"text": "(Rose et al., 1990)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "System Output:",
	"sec_num": "5.1.3"
	},
	{
	"text": "System output for example paper.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 12",
	"sec_num": null
	},
	{
	"text": "formation about research goals of the paper: Sentence 41 states the goal in explicit terms, but it also contains a contrastive statement, which the annotator decided to rate higher than the goal statement. Both sentences 12 and 150 give high-level descriptions of the work that might pass as a goal statement. Similarly, in sentence 21 the agent and action features detected that the first part of the sentence has something to do with comparing methods, and the system then (plausibly but incorrectly) decided to classify the sentence as CONTRAST. All in all, we feel that the extracted material conveys the rhetorical status adequately. An extrinsic evaluation additionally showed that the end result provides considerable added value when compared to sentence extracts (Teufel 2001) .",
	"cite_spans": [
	{
	"start": 772,
	"end": 785,
	"text": "(Teufel 2001)",
	"ref_id": "BIBREF76"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Figure 12",
	"sec_num": null
	},
	{
	"text": "The classifier for rhetorical status that we evaluated in the previous section is an important first step in our implementation; the next step is the determination of relevant sentences in the text. One simple solution for relevance decision would be to use all AIM, BASIS, and CONTRAST sentences, as these categories are rare overall. The classifier we use has the nice property of roughly keeping the distribution of target categories, so that we end up with a sensible number of these sentences. The strategy of using all AIM, CONTRAST, and BASIS sentences can be evaluated in a similar vein to the previous experiment. In terms of relevance, the asterisk in figure 12 marks sentences that the human judge found particularly relevant in the overall context (cf. the full set in figure 5 ). Six out of all 15 sentences, and 6 out of the 10 sentences that received the correct rhetorical status, were judged relevant in the example. Table 12 reports the figure for the entire corpus by comparing the system's output of correctly classified rhetorical categories to human judgment. In all cases, the results are far above the nontrivial baseline. On AIM, CONTRAST, and BASIS sentences, our system achieves very high precision values of 96%, 70%, and 71%. Recall is lower at 70%, 24%, and 39%, but low recall is less of a problem in our final task. Therefore, the main bottleneck is correct rhetorical classification. Once that is accomplished, the selected categories show high agreement with human judgment and should therefore represent good material for further processing steps.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 781,
	"end": 789,
	"text": "figure 5",
	"ref_id": null
	},
	{
	"start": 934,
	"end": 942,
	"text": "Table 12",
	"ref_id": "TABREF0"
	}
	],
	"eq_spans": [],
	"section": "Relevance Determination",
	"sec_num": "5.2"
	},
	{
	"text": "If, however, one is also interested in selecting BACKGROUND sentences, as we are, simply choosing all BACKGROUND sentences would result in low precision of 16% (albeit with a high recall of 83%), which does not seem to be the optimal solution. We therefore use a second classifier for finding the most relevant sentences independently that was trained on the relevance gold standard. Our best classifier operates at a precision of 46.5% and recall of 45.2% (using the features Location, Section Struct, Paragraph Struct, Title, TF*IDF, Formulaic, and Citation for classification). The second classifier (cf. rightmost columns in figure 12) raises the precision for BACKGROUND sentences from 16% to 38%, while keeping recall high at 88%. This example shows that the right procedure for relevance determination changes from category to category and also depends on the final task one is trying to accomplish.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Relevance Determination",
	"sec_num": "5.2"
	},
	{
	"text": "Relevance by human selection: Precision (P) and recall (R). ",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Table 12",
	"sec_num": null
	},
	{
	"text": "We have presented a new method for content selection from scientific articles. The analysis is genre-specific; it is based on rhetorical phenomena specific to academic writing, such as problem-solution structure, explicit intellectual attribution, and statements of relatedness to other work. The goal of the analysis is to identify the contribution of an article in relation to background material and to other specific current work. Our methodology is situated between text extraction methods and fact extraction (template-filling) methods: Although our analysis has the advantage of being more context-sensitive than text extraction methods, it retains the robustness of this approach toward different subdomains, presentational traditions, and writing styles.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Contribution",
	"sec_num": "6.1"
	},
	{
	"text": "Like fact extraction methods (e.g., Radev and McKeown 1998) , our method also uses a \"template\" whose slots are being filled during analysis. The slots of our template are defined as rhetorical categories (like \"Contrast\") rather than by domain-specific categories (like \"Perpetrator\"). This makes it possible for our approach to deal with texts of different domains and unexpected topics.",
	"cite_spans": [
	{
	"start": 36,
	"end": 59,
	"text": "Radev and McKeown 1998)",
	"ref_id": "BIBREF62"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Contribution",
	"sec_num": "6.1"
	},
	{
	"text": "Sparck Jones (1999) argues that it is crucial for a summarization strategy to relate the large-scale document structure of texts to readers' tasks in the real world (i.e., to the proposed use of the summaries). We feel that incorporating a robust analysis of discourse structure into a document summarizer is one step along this way.",
	"cite_spans": [
	{
	"start": 7,
	"end": 19,
	"text": "Jones (1999)",
	"ref_id": "BIBREF73"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Contribution",
	"sec_num": "6.1"
	},
	{
	"text": "Our practical contributions are twofold. First, we present a scheme for the annotation of sentences with rhetorical status, and we have shown that the annotation is stable (K = .82, .81, .76) and reproducible (K = .71). Since these results indicate that the annotation is reliable, we use it as our gold standard for evaluation and training.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Contribution",
	"sec_num": "6.1"
	},
	{
	"text": "Second, we present a machine learning system for the classification of sentences by relevance and by rhetorical status. The contribution here is not the statistical classifier, which is well-known and has been used in a similar task by Kupiec, Pedersen, and Oren (1995) , but instead the features we use. We have adapted 13 sentential features in such a way that they work robustly for our task (i.e., for unrestricted, real-world text). We also present three new features that detect scientific metadiscourse in a novel way. The results of an intrinsic system evaluation show that the system can identify sentences expressing the specific goal of a paper with 57% precision and 79% recall, sentences expressing criticism or contrast with 57% precision and 42% recall, and sentences expressing a continuation relationship to other work with 62% precision and 43% recall. This substantially improves a baseline of text classification which uses only a TF*IDF model over words. The agreement of correctly identified rhetorical roles with human relevance judgments is even higher (96% precision and 70% recall for goal statements, 70% precision and 24% recall for contrast, 71% precision and 39% recall for continuation). We see these results as an indication that shallow discourse processing with a well-designed set of surface-based indicators is possible.",
	"cite_spans": [
	{
	"start": 236,
	"end": 269,
	"text": "Kupiec, Pedersen, and Oren (1995)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Contribution",
	"sec_num": "6.1"
	},
	{
	"text": "The metadiscourse features, one focus of our work, currently depend on manual resources. The experiments reported here explore whether metadiscourse information is useful for the automatic determination of rhetorical status (as opposed to more shallow features), and this is clearly the case. The next step, however, should be the automatic creation of such resources. For the task of dialogue act disambiguation, Samuel, Carberry, and Vijay-Shanker (1999) suggest a method of automatically finding cue phrases for disambiguation. It may be possible to apply this or a similar method to our data and to compare the performance of automatically gained resources with manual ones.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Limitations and Future Work",
	"sec_num": "6.2"
	},
	{
	"text": "Further work can be done on the semantic verb clusters described in section 4.2. Klavans and Kan (1998) , who use verb clusters for document classification according to genre, observe that verb information is rarely used in current practical natural language applications. Most tasks such as information extraction and document classification identify and use nominal constructs instead (e.g., noun phrases, TF*IDF words and phrases).",
	"cite_spans": [
	{
	"start": 81,
	"end": 103,
	"text": "Klavans and Kan (1998)",
	"ref_id": "BIBREF39"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Limitations and Future Work",
	"sec_num": "6.2"
	},
	{
	"text": "The verb clusters we employ were created using our intuition of which type of verb similarity would be useful in the genre and for the task. There are good reasons for using such a hand-crafted, genre-specific verb lexicon instead of a general resource such as WordNet or Levin's (1993) classes: Many verbs used in the domain of scientific argumentation have assumed a specialized meaning, which our lexicon readily encodes. Klavans and Kan's classes, which are based on Levin's classes, are also manually created. Resnik and Diab (2000) present yet other measures of verb similarity, which could be used to arrive at a more data-driven definition of verb classes. We are currently comparing our verb clusterings to Klavans and Kan's, and to bottom-up clusters of verb similarities generated from our annotated data.",
	"cite_spans": [
	{
	"start": 272,
	"end": 286,
	"text": "Levin's (1993)",
	"ref_id": "BIBREF46"
	},
	{
	"start": 515,
	"end": 537,
	"text": "Resnik and Diab (2000)",
	"ref_id": "BIBREF64"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Limitations and Future Work",
	"sec_num": "6.2"
	},
	{
	"text": "The recognition of agents, which is already the second-best feature in the pool, could be further improved by including named entity recognition and anaphora resolution. Named entity recognition would help in cases like the following, LHIP provides a processing method which allows selected portions of the input to be ignored or handled differently. (S-5, 9408006) where LHIP is the name of the authors' approach and should thus be tagged as US AGENT; to do so, however, one would need to recognize it as a named approach, which is associated with the authors. It is very likely that such a treatment, which would have to include information from elsewhere in the text, would improve results, particularly as named approaches are frequent in the computational linguistics domain. Information about named approaches in themselves would also be an important aspect to include in summaries or citation indexes. Anaphora resolution helps in cases in which the agent is syntactically ambiguous between own and other approaches (e.g., this system). To test whether and how much performance would improve, we manually simulated anaphora resolution on the 632 occurrences of REF AGENT in the development corpus. (In the experiments in section 5 these occurrences had been excluded from the Agent feature by giving them the value None; we include them now in their disambiguated state). Of the 632 REF AGENTs, 436 (69%) were classified as US AGENT, 175 (28%) as THEM AGENT, and 20 (3%) as GENERAL AGENT. As a result of this manual disambiguation, the performance of the Agent feature increased dramatically from K = .08 to K = .14 and that of SegAgent from K = .19 to K = .22. This is a clear indication of the potential added value of anaphora resolution for our task.",
	"cite_spans": [
	{
	"start": 351,
	"end": 365,
	"text": "(S-5, 9408006)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Limitations and Future Work",
	"sec_num": "6.2"
	},
	{
	"text": "As far as the statistical classification is concerned, our results are still far from perfect. Obvious ways of improving performance are the use of a more sophisticated statistical classifier and more training material. We have experimented with a maximum entropy model, Repeated Incremental Pruning to Produce Error Reduction (RIP-PER), and decision trees; preliminary results do not show significant improvement over the na\u00efve Bayesian model. One problem is that 4% of the sentences in our current annotated material are ambiguous: They receive the same feature representation but are classified differently by the annotators. A possible solution is to find better and more distinctive features; we believe that robust, higher-level features like actions and agents are a step in the right direction. We also suspect that a big improvement",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Limitations and Future Work",
	"sec_num": "6.2"
	}
	],
	"back_matter": [
	{
	"text": "The work reported in this article was conducted while both authors were in the HCRC Language Technology Group at the University of Edinburgh.The authors would like to thank Jean Carletta for her help with the experimental design, Chris Brew for many helpful discussions, Claire Grover and Andrei Mikheev for advice on the XML implementation, and the annotators, Vasilis Karaiskos and Anne Wilson, for their meticulous work and criticism, which led to several improvements in the annotation scheme. Thanks also to Byron Georgantopolous, who helped to collect the first version of the corpus, and to the four anonymous reviewers.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Acknowledgments",
	"sec_num": null
	},
	{
	"text": "could be achieved with smaller annotation units. Many errors come from instances in which one half of a sentence serves one rhetorical purpose, the other another, as in the following example:The current paper shows how to implement this general notion, without following Krifka's analysis in detail. (S-10, 9411019) Here, the first part describes the paper's research goal, whereas the second expresses a contrast. Currently, one target category needs to be associated with the whole sentence (according to a rule in the guidelines, AIM is given preference over CONTRAST). As an undesired side effect, the CONTRAST-like textual parts (and the features associated with this text piece, e.g., the presence of an author's name) are wrongly associated with the AIM target category. If we allowed for a smaller annotation unit (e.g., at the clause level), this systematic noise in the training data could be removed. Another improvement in classification accuracy might be achieved by performing the classification in a cascading way. The system could first perform a classification into OWN-like classes (OWN, AIM, and TEXTUAL pooled), OTHER-like categories (OTHER, CONTRAST, and BASIS pooled), and BACKGROUND, similar to the way human annotation proceeds. Subclassification among these classes would then lead to the final seven-way classification.",
	"cite_spans": [
	{
	"start": 300,
	"end": 315,
	"text": "(S-10, 9411019)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "annex",
	"sec_num": null
	},
	{
	"text": "No.Title, Conference, Authors",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Appendix: List of articles in CL development corpus",
	"sec_num": null
	}
	],
	"bib_entries": {
	"BIBREF0": {
	"ref_id": "b0",
	"title": "Estimation of Word Cooccurrence Probabilities (ACL94)",
	"authors": [
	{
	"first": "I",
	"middle": [],
	"last": "Dagan",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Similarity-Based Estimation of Word Cooccurrence Probabilities (ACL94), I. Dagan et al.",
	"links": null
	},
	"BIBREF1": {
	"ref_id": "b1",
	"title": "Temporal Relations: Reference or Discourse Coherence? (ACL94 Student), A. Kehler 2 9405004 Syntactic-Head-Driven Generation (COLING94), E. Koenig",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Temporal Relations: Reference or Discourse Coherence? (ACL94 Student), A. Kehler 2 9405004 Syntactic-Head-Driven Generation (COLING94), E. Koenig",
	"links": null
	},
	"BIBREF2": {
	"ref_id": "b2",
	"title": "Common Topics and Coherent Situations: Interpreting Ellipsis in the Context of Discourse Inference (ACL94)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Common Topics and Coherent Situations: Interpreting Ellipsis in the Context of Dis- course Inference (ACL94), A. Kehler",
	"links": null
	},
	"BIBREF4": {
	"ref_id": "b4",
	"title": "Grammar Specialization through Entropy Thresholds (ACL94)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Grammar Specialization through Entropy Thresholds (ACL94), C. Samuelsson",
	"links": null
	},
	"BIBREF5": {
	"ref_id": "b5",
	"title": "An Integrated Heuristic Scheme for Partial Parse Evaluation (ACL94 Student), A. Lavie 7 9405028 Semantics of Complex Sentences in Japanese (COLING94), H. Nakagawa, S. Nishizawa 8 9405033 Relating Complexity to Practical Performance in Parsing with Wide-Coverage Unification Grammars (ACL94)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "An Integrated Heuristic Scheme for Partial Parse Evaluation (ACL94 Student), A. Lavie 7 9405028 Semantics of Complex Sentences in Japanese (COLING94), H. Nakagawa, S. Nishizawa 8 9405033 Relating Complexity to Practical Performance in Parsing with Wide-Coverage Unifica- tion Grammars (ACL94), J. Carroll",
	"links": null
	},
	"BIBREF6": {
	"ref_id": "b6",
	"title": "Dual-Coding Theory and Connectionist Lexical Selection (ACL94 Student), Y. Wang",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "10",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Dual-Coding Theory and Connectionist Lexical Selection (ACL94 Student), Y. Wang 10 9407011",
	"links": null
	},
	"BIBREF7": {
	"ref_id": "b7",
	"title": "Allen 11 9408003 Typed Feature Structures as Descriptions (COLING94 Reserve), P. King 12 9408004 Parsing with Principles and Probabilities (ACL94 Workshop), A. Fordham, M. Crocker 13 9408006 LHIP: Extended DCGs for Configurable Robust Parsing (COLING94), A. Ballim, G. Russell 14 9408011 Distributional Clustering of English Words (ACL93)",
	"authors": [
	{
	"first": "D",
	"middle": [],
	"last": "Traum",
	"suffix": ""
	},
	{
	"first": "J",
	"middle": [
	"; F"
	],
	"last": "Pereira",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Discourse Obligations in Dialogue Processing (ACL94), D. Traum, J. Allen 11 9408003 Typed Feature Structures as Descriptions (COLING94 Reserve), P. King 12 9408004 Parsing with Principles and Probabilities (ACL94 Workshop), A. Fordham, M. Crocker 13 9408006 LHIP: Extended DCGs for Configurable Robust Parsing (COLING94), A. Bal- lim, G. Russell 14 9408011 Distributional Clustering of English Words (ACL93), F. Pereira et al.",
	"links": null
	},
	"BIBREF8": {
	"ref_id": "b8",
	"title": "Qualitative and Quantitative Models of Speech Translation (ACL94 Workshop)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Qualitative and Quantitative Models of Speech Translation (ACL94 Workshop), H. Alshawi",
	"links": null
	},
	"BIBREF9": {
	"ref_id": "b9",
	"title": "9410006 Evaluating Discourse Processing Algorithms (ACL89), M. Walker 20 9410008 Recognizing Text Genres with Simple Metrics Using Discriminant Analysis (COLING94), J. Karlgren, D. Cutting 21 9410009 Reserve Lexical Functions and Machine Translation (COLING94), D. Heylen et al. 22 9410012 Does Baum-Welch Re-estimation Help Taggers? (ANLP94), D. Elworthy 23 9410022 Automated Tone Transcription (ACL94 SIG), S. Bird 24 9410032 Planning Argumentative Texts (COLING94), X. Huang 25 9410033 Default Handling in Incremental Generation (COLING94)",
	"authors": [
	{
	"first": "S",
	"middle": [],
	"last": "Brennan",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "27",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "An Experiment on Learning Appropriate Selectional Restrictions from a Parsed Corpus (COLING94), F. Ribas 17 9410001 Improving Language Models by Clustering Training Sentences (ANLP94), D. Carter 18 9410005 A Centering Approach to Pronouns (ACL87), S. Brennan et al. 19 9410006 Evaluating Discourse Processing Algorithms (ACL89), M. Walker 20 9410008 Recognizing Text Genres with Simple Metrics Using Discriminant Analysis (COLING94), J. Karlgren, D. Cutting 21 9410009 Reserve Lexical Functions and Machine Translation (COLING94), D. Heylen et al. 22 9410012 Does Baum-Welch Re-estimation Help Taggers? (ANLP94), D. Elworthy 23 9410022 Automated Tone Transcription (ACL94 SIG), S. Bird 24 9410032 Planning Argumentative Texts (COLING94), X. Huang 25 9410033 Default Handling in Incremental Generation (COLING94), K. Harbusch et al. 26 9411019 Focus on \"Only\" and \"Not\" (COLING94), A. Ramsay 27 9411021",
	"links": null
	},
	"BIBREF10": {
	"ref_id": "b10",
	"title": "CUG on Chemical Abstract Machine (COLING94), S. Tojo",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "28",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Free-Ordered CUG on Chemical Abstract Machine (COLING94), S. Tojo 28 9411023",
	"links": null
	},
	"BIBREF12": {
	"ref_id": "b12",
	"title": "Segmenting Speech without a Lexicon: The Roles of Phonotactics and Speech Source (ACL94 SIG), T. Cartwright, M. Brent 30 9412008 Analysis of Japanese Compound Nouns Using Collocational Information (COLING94)",
	"authors": [
	{
	"first": "Y",
	"middle": [],
	"last": "Kobayasi",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Segmenting Speech without a Lexicon: The Roles of Phonotactics and Speech Source (ACL94 SIG), T. Cartwright, M. Brent 30 9412008 Analysis of Japanese Compound Nouns Using Collocational Information (COLING94), Y. Kobayasi et al.",
	"links": null
	},
	"BIBREF13": {
	"ref_id": "b13",
	"title": "Bottom-Up Earley Deduction (COLING94), G. Erbach 32 9502005 Off-Line Optimization for Earley-Style HPSG Processing (EACL95), G. Minnen et al. 33 9502006 Rapid Development of Morphological Descriptions for Full Language Processing Systems (EACL95), D. Carter 34 9502009 On Learning More Appropriate Selectional Restrictions (EACL95), F. Ribas 35 9502014 Ellipsis and Quantification: A Substitutional Approach (EACL95), R. Crouch 36 9502015 The Semantics of Resource Sharing in Lexical-Functional Grammar (EACL95)",
	"authors": [
	{
	"first": "A",
	"middle": [],
	"last": "Kehler",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Bottom-Up Earley Deduction (COLING94), G. Erbach 32 9502005 Off-Line Optimization for Earley-Style HPSG Processing (EACL95), G. Minnen et al. 33 9502006 Rapid Development of Morphological Descriptions for Full Language Processing Sys- tems (EACL95), D. Carter 34 9502009 On Learning More Appropriate Selectional Restrictions (EACL95), F. Ribas 35 9502014 Ellipsis and Quantification: A Substitutional Approach (EACL95), R. Crouch 36 9502015 The Semantics of Resource Sharing in Lexical-Functional Grammar (EACL95), A. Kehler et al.",
	"links": null
	},
	"BIBREF14": {
	"ref_id": "b14",
	"title": "Algorithms for Analysing the Temporal Structure of Discourse (EACL95)",
	"authors": [
	{
	"first": "J",
	"middle": [],
	"last": "Hitzeman",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Algorithms for Analysing the Temporal Structure of Discourse (EACL95), J. Hitzeman et al.",
	"links": null
	},
	"BIBREF15": {
	"ref_id": "b15",
	"title": "42 9502031 Cooperative Error Handling and Shallow Processing (EACL95 Student), T. Bowden 43 9502033 An Algorithm to Co-ordinate Anaphora Resolution and PPS Disambiguation Process (EACL95 Student), S. Azzam 44 9502035 Incorporating \"Unconscious Reanalysis\" into an Incremental, Monotonic Parser (EACL95 Student), P. Sturt 45 9502037 A State-Transition Grammar for Data-Oriented Parsing (EACL95 Student), D. Tugwell 46 9502038 Implementation and Evaluation of a German HMM for POS Disambiguation (EACL95 Workshop), H. Feldweg 47 9502039 Multilingual Sentence Categorization According to Language (EACL95 Workshop",
	"authors": [
	{
	"first": "B",
	"middle": [],
	"last": "Keller",
	"suffix": ""
	},
	{
	"first": "D",
	"middle": [
	";"
	],
	"last": "",
	"suffix": ""
	},
	{
	"first": "K",
	"middle": [],
	"last": "Lee",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "Giguet 48 9503002 Computational Dialectology in Irish Gaelic (EACL95), B. Kessler 49 9503004 Creating a Tagset, Lexicon and Guesser for a French Tagger (EACL95 Workshop), J. Chanod, P. Tapanainen 50 9503005 A Specification Language for Lexical Functional Grammars (EACL95), P. Blackburn, C. Gardent 51 9503007 The Semantics of Motion (EACL95), P. Sablayrolles 52 9503009 Distributional Part-of-Speech Tagging (EACL95), H. Schuetze 53 9503013 Incremental Interpretation: Applications, Theory, and Relationship to Dynamic Semantics (COLING95), D. Milward, R. Cooper 54 9503014 Non-constituent Coordination: Theory and Practice (COLING94), D",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "A Tractable Extension of Linear Indexed Grammars (EACL95), B. Keller, D. Weir 39 9502022 Stochastic HPSG (EACL95), C. Brew 40 9502023 Splitting the Reference Time: Temporal Anaphora and Quantification in DRT (EACL95), R. Nelken, N. Francez 41 9502024 A Robust Parser Based on Syntactic Information (EACL95), K. Lee et al. 42 9502031 Cooperative Error Handling and Shallow Processing (EACL95 Student), T. Bowden 43 9502033 An Algorithm to Co-ordinate Anaphora Resolution and PPS Disambiguation Process (EACL95 Student), S. Azzam 44 9502035 Incorporating \"Unconscious Reanalysis\" into an Incremental, Monotonic Parser (EACL95 Student), P. Sturt 45 9502037 A State-Transition Grammar for Data-Oriented Parsing (EACL95 Student), D. Tugwell 46 9502038 Implementation and Evaluation of a German HMM for POS Disambiguation (EACL95 Workshop), H. Feldweg 47 9502039 Multilingual Sentence Categorization According to Language (EACL95 Workshop), E. Giguet 48 9503002 Computational Dialectology in Irish Gaelic (EACL95), B. Kessler 49 9503004 Creating a Tagset, Lexicon and Guesser for a French Tagger (EACL95 Workshop), J. Chanod, P. Tapanainen 50 9503005 A Specification Language for Lexical Functional Grammars (EACL95), P. Blackburn, C. Gardent 51 9503007 The Semantics of Motion (EACL95), P. Sablayrolles 52 9503009 Distributional Part-of-Speech Tagging (EACL95), H. Schuetze 53 9503013 Incremental Interpretation: Applications, Theory, and Relationship to Dynamic Seman- tics (COLING95), D. Milward, R. Cooper 54 9503014 Non-constituent Coordination: Theory and Practice (COLING94), D. Milward 55 9503015 Incremental Interpretation of Categorial Grammar (EACL95), D. Milward 56 9503017 Redundancy in Collaborative Dialogue (COLING92), M. Walker 57 9503018 Discourse and Deliberation: Testing a Collaborative Strategy (COLING94), M. Walker 58 9503023 A Fast Partial Parse of Natural Language Sentences Using a Connectionist Method (EACL95), C. Lyon, B. Dickerson",
	"links": null
	},
	"BIBREF16": {
	"ref_id": "b16",
	"title": "Distance Vectors from Dictionaries (COLING94)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Occurrence Vectors from Corpora vs. Distance Vectors from Dictionaries (COLING94), Y. Niwa, Y. Nitta",
	"links": null
	},
	"BIBREF17": {
	"ref_id": "b17",
	"title": "Stenton 62 9504007 Mixed Initiative in Dialogue: An Investigation into Discourse Segmentation (ACL90), M. Walker, S. Whittaker 63 9504017 A Uniform Treatment of Pragmatic Inferences in Simple and Complex Utterances and Sequences of Utterances (ACL95), D. Marcu, G. Hirst 64 9504024 A Morphographemic Model for Error Correction in Nonconcatenative Strings (ACL95), T",
	"authors": [
	{
	"first": "S",
	"middle": [],
	"last": "Whittaker",
	"suffix": ""
	},
	{
	"first": "P",
	"middle": [],
	"last": "",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Tagset Design and Inflected Languages (EACL95 Workshop), D. Elworthy 61 9504006 Cues and Control in Expert-Client Dialogues (ACL88), S. Whittaker, P. Stenton 62 9504007 Mixed Initiative in Dialogue: An Investigation into Discourse Segmentation (ACL90), M. Walker, S. Whittaker 63 9504017 A Uniform Treatment of Pragmatic Inferences in Simple and Complex Utterances and Sequences of Utterances (ACL95), D. Marcu, G. Hirst 64 9504024 A Morphographemic Model for Error Correction in Nonconcatenative Strings (ACL95), T. Bowden, G. Kiraz",
	"links": null
	},
	"BIBREF18": {
	"ref_id": "b18",
	"title": "The Intersection of Finite State Automata and Definite Clause Grammars (ACL95)",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "The Intersection of Finite State Automata and Definite Clause Grammars (ACL95), G. van Noord",
	"links": null
	},
	"BIBREF19": {
	"ref_id": "b19",
	"title": "67 9504030 Statistical Decision-Tree Models for Parsing (ACL95), D. Magerman 68 9504033 Corpus Statistics Meet the Noun Compound: Some Empirical Results (ACL95), M. Lauer 79 9504034 Bayesian Grammar Induction for Language Modeling (ACL95), S. Chen 70 9505001 Response Generation in Collaborative Negotiation (ACL95)",
	"authors": [
	{
	"first": "V",
	"middle": [],
	"last": "Poznanski",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "An Efficient Generation Algorithm for Lexicalist MT (ACL95)",
	"volume": "72",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "An Efficient Generation Algorithm for Lexicalist MT (ACL95), V. Poznanski et al. 67 9504030 Statistical Decision-Tree Models for Parsing (ACL95), D. Magerman 68 9504033 Corpus Statistics Meet the Noun Compound: Some Empirical Results (ACL95), M. Lauer 79 9504034 Bayesian Grammar Induction for Language Modeling (ACL95), S. Chen 70 9505001 Response Generation in Collaborative Negotiation (ACL95), J. Chu-Carroll, S. Carberry 71 9506004 Using Higher-Order Logic Programming for Semantic Interpretation of Coordinate Con- structs (ACL95), S. Kulick 72 9511001",
	"links": null
	},
	"BIBREF20": {
	"ref_id": "b20",
	"title": "Countability and Number in Japanese-to-English Machine Translation (COLING94)",
	"authors": [
	{
	"first": "F",
	"middle": [],
	"last": "Bond",
	"suffix": ""
	}
	],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Countability and Number in Japanese-to-English Machine Translation (COLING94), F. Bond et al.",
	"links": null
	},
	"BIBREF21": {
	"ref_id": "b21",
	"title": "The rules behind roles",
	"authors": [],
	"year": 2000,
	"venue": "Disambiguating Noun Groupings with Respect to WordNet Senses (ACL95 Workshop), P. Resnik 74 9601004 Similarity between Words Computed by Spreading Activation on an English Dictionary (EACL93), H. Kozima, T. Furugori 75 9604019 Magic for Filter Optimization in Dynamic Bottom-up Processing (ACL96), G. Minnen 76 9604022 Unsupervised Learning of Word-Category Guessing Rules (ACL96), A. Mikheev 77 9605013 Learning Dependencies between Case Frame Slots (COLING96), H. Li, N. Abe 78 9605014 Clustering Words with the MDL Principle (COLING96), H. Li, N. Abe 79 9605016 Parsing for Semidirectional Lambek Grammar is NP-Complete (ACL96)",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Disambiguating Noun Groupings with Respect to WordNet Senses (ACL95 Workshop), P. Resnik 74 9601004 Similarity between Words Computed by Spreading Activation on an English Dictionary (EACL93), H. Kozima, T. Furugori 75 9604019 Magic for Filter Optimization in Dynamic Bottom-up Processing (ACL96), G. Minnen 76 9604022 Unsupervised Learning of Word-Category Guessing Rules (ACL96), A. Mikheev 77 9605013 Learning Dependencies between Case Frame Slots (COLING96), H. Li, N. Abe 78 9605014 Clustering Words with the MDL Principle (COLING96), H. Li, N. Abe 79 9605016 Parsing for Semidirectional Lambek Grammar is NP-Complete (ACL96), J. Doerre References Barzilay, Regina, Michael Collins, Julia Hirschberg, and Steve Whittaker. 2000. The rules behind roles. In Proceedings of AAAI-00.",
	"links": null
	},
	"BIBREF22": {
	"ref_id": "b22",
	"title": "Information fusion in the context of multi-document summarization",
	"authors": [
	{
	"first": "Regina",
	"middle": [],
	"last": "Barzilay",
	"suffix": ""
	},
	{
	"first": "Kathleen",
	"middle": [
	"R"
	],
	"last": "Mckeown",
	"suffix": ""
	},
	{
	"first": "Michael",
	"middle": [],
	"last": "Elhadad",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics (ACL-99)",
	"volume": "",
	"issue": "",
	"pages": "550--557",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Barzilay, Regina, Kathleen R. McKeown, and Michael Elhadad. 1999. Information fusion in the context of multi-document summarization. In Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics (ACL-99), pages 550-557.",
	"links": null
	},
	"BIBREF23": {
	"ref_id": "b23",
	"title": "Man-made index for technical literature-An experiment",
	"authors": [
	{
	"first": "Phyllis",
	"middle": [
	"B"
	],
	"last": "Baxendale",
	"suffix": ""
	}
	],
	"year": 1958,
	"venue": "IBM Journal of Research and Development",
	"volume": "2",
	"issue": "4",
	"pages": "354--361",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Baxendale, Phyllis B. 1958. Man-made index for technical literature-An experiment. IBM Journal of Research and Development, 2(4):354-361.",
	"links": null
	},
	"BIBREF24": {
	"ref_id": "b24",
	"title": "Automatic condensation of electronic publications by sentence selection",
	"authors": [
	{
	"first": "Douglas",
	"middle": [],
	"last": "Biber",
	"suffix": ""
	},
	{
	"first": "Ronald",
	"middle": [],
	"last": "Brandow",
	"suffix": ""
	},
	{
	"first": "Karl",
	"middle": [],
	"last": "Mitze",
	"suffix": ""
	},
	{
	"first": "Lisa",
	"middle": [
	"F"
	],
	"last": "Rau",
	"suffix": ""
	}
	],
	"year": 1995,
	"venue": "Information Processing and Management",
	"volume": "31",
	"issue": "5",
	"pages": "675--685",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Biber, Douglas. 1995. Dimensions of Register Variation: A Cross-Linguistic Comparison. Cambridge University Press, Cambridge, England. Brandow, Ronald, Karl Mitze, and Lisa F. Rau. 1995. Automatic condensation of electronic publications by sentence selection. Information Processing and Management, 31(5):675-685.",
	"links": null
	},
	"BIBREF25": {
	"ref_id": "b25",
	"title": "Assessing agreement on classification tasks. The kappa statistic",
	"authors": [
	{
	"first": "Jean",
	"middle": [],
	"last": "Carletta",
	"suffix": ""
	}
	],
	"year": 1996,
	"venue": "Computational Linguistics",
	"volume": "22",
	"issue": "2",
	"pages": "249--254",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Carletta, Jean. 1996. Assessing agreement on classification tasks. The kappa statistic. Computational Linguistics, 22(2):249-254.",
	"links": null
	},
	"BIBREF26": {
	"ref_id": "b26",
	"title": "Advances in domain independent linear text segmentation",
	"authors": [
	{
	"first": "Freddy",
	"middle": [
	"Y Y"
	],
	"last": "Choi",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Proceedings of the Sixth Applied Natural Language Conference (ANLP-00) and the First Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL-00)",
	"volume": "",
	"issue": "",
	"pages": "26--33",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Choi, Freddy Y. Y. 2000. Advances in domain independent linear text segmentation. In Proceedings of the Sixth Applied Natural Language Conference (ANLP-00) and the First Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL-00), pages 26-33.",
	"links": null
	},
	"BIBREF27": {
	"ref_id": "b27",
	"title": "The computation and language e-print archive",
	"authors": [
	{
	"first": "",
	"middle": [],
	"last": "Cmp Lg",
	"suffix": ""
	}
	],
	"year": 1994,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "CMP LG. 1994. The computation and language e-print archive. http://xxx.lanl.gov/cmp-lg.",
	"links": null
	},
	"BIBREF28": {
	"ref_id": "b28",
	"title": "Accurate methods for the statistics of surprise and coincidence",
	"authors": [
	{
	"first": "Ted",
	"middle": [],
	"last": "Dunning",
	"suffix": ""
	}
	],
	"year": 1993,
	"venue": "Computational Linguistics",
	"volume": "19",
	"issue": "1",
	"pages": "61--74",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Dunning, Ted. 1993. Accurate methods for the statistics of surprise and coincidence. Computational Linguistics, 19(1):61-74.",
	"links": null
	},
	"BIBREF29": {
	"ref_id": "b29",
	"title": "New methods in automatic extracting",
	"authors": [
	{
	"first": "H",
	"middle": [
	"P"
	],
	"last": "Edmundson",
	"suffix": ""
	}
	],
	"year": 1969,
	"venue": "Journal of the Association for Computing Machinery",
	"volume": "16",
	"issue": "2",
	"pages": "264--285",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Edmundson, H. P. 1969. New methods in automatic extracting. Journal of the Association for Computing Machinery, 16(2):264-285.",
	"links": null
	},
	"BIBREF30": {
	"ref_id": "b30",
	"title": "Automatic abstracting and indexing-Survey and recommendations",
	"authors": [
	{
	"first": "H",
	"middle": [
	"P"
	],
	"last": "Edmundson",
	"suffix": ""
	},
	{
	"first": "R",
	"middle": [
	"E"
	],
	"last": "Wyllys",
	"suffix": ""
	}
	],
	"year": 1961,
	"venue": "Communications of the ACM",
	"volume": "4",
	"issue": "5",
	"pages": "226--234",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Edmundson, H. P., and Wyllys, R. E. 1961. Automatic abstracting and indexing-Survey and recommendations. Communications of the ACM, 4(5):226-234.",
	"links": null
	},
	"BIBREF31": {
	"ref_id": "b31",
	"title": "Citation Indexing: Its Theory and Application in Science, Technology and Humanities",
	"authors": [
	{
	"first": "Eugene",
	"middle": [],
	"last": "Garfield",
	"suffix": ""
	}
	],
	"year": 1979,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Garfield, Eugene. 1979. Citation Indexing: Its Theory and Application in Science, Technology and Humanities. J. Wiley, New York.",
	"links": null
	},
	"BIBREF32": {
	"ref_id": "b32",
	"title": "Producing intelligent telegraphic text reduction to provide an audio scanning service for the blind",
	"authors": [
	{
	"first": "Gregory",
	"middle": [],
	"last": "Grefenstette",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Working Notes of the AAAI Spring Symposium on Intelligent Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "111--117",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Grefenstette, Gregory. 1998. Producing intelligent telegraphic text reduction to provide an audio scanning service for the blind. In D. R. Radev and E. H. Hovy, editors, Working Notes of the AAAI Spring Symposium on Intelligent Text Summarization, pages 111-117.",
	"links": null
	},
	"BIBREF33": {
	"ref_id": "b33",
	"title": "LT TTT version 1.0: Text tokenisation software",
	"authors": [
	{
	"first": "Claire",
	"middle": [],
	"last": "Grover",
	"suffix": ""
	},
	{
	"first": "Andrei",
	"middle": [],
	"last": "Mikheev",
	"suffix": ""
	},
	{
	"first": "Colin",
	"middle": [],
	"last": "Matheson",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Grover, Claire, Andrei Mikheev, and Colin Matheson. 1999. LT TTT version 1.0: Text tokenisation software. Technical Report, Human Communication Research Centre, University of Edinburgh. http://www.ltg.ed.ac.uk/software/ttt/.",
	"links": null
	},
	"BIBREF34": {
	"ref_id": "b34",
	"title": "Texttiling: Segmenting text into multi-paragraph subtopic passages",
	"authors": [
	{
	"first": "Marti",
	"middle": [
	"A"
	],
	"last": "Hearst",
	"suffix": ""
	}
	],
	"year": 1997,
	"venue": "Computational Linguistics",
	"volume": "23",
	"issue": "1",
	"pages": "33--64",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Hearst, Marti A. 1997. Texttiling: Segmenting text into multi-paragraph subtopic passages. Computational Linguistics, 23(1):33-64.",
	"links": null
	},
	"BIBREF35": {
	"ref_id": "b35",
	"title": "Persuasion and context: The pragmatics of academic metadiscourse",
	"authors": [
	{
	"first": "Ken",
	"middle": [],
	"last": "Hyland",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Journal of Pragmatics",
	"volume": "30",
	"issue": "4",
	"pages": "437--455",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Hyland, Ken. 1998. Persuasion and context: The pragmatics of academic metadiscourse. Journal of Pragmatics, 30(4):437-455.",
	"links": null
	},
	"BIBREF36": {
	"ref_id": "b36",
	"title": "Summarization evaluation methods: Experiments and analysis",
	"authors": [
	{
	"first": "Hongyan",
	"middle": [],
	"last": "Jing",
	"suffix": ""
	},
	{
	"first": "Regina",
	"middle": [],
	"last": "Barzilay",
	"suffix": ""
	},
	{
	"first": "Kathleen",
	"middle": [
	"R"
	],
	"last": "Mckeown",
	"suffix": ""
	},
	{
	"first": "Michael",
	"middle": [],
	"last": "Elhadad",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Working Notes of the AAAI Spring Symposium on Intelligent Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "60--68",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Jing, Hongyan, Regina Barzilay, Kathleen R. McKeown, and Michael Elhadad. 1998. Summarization evaluation methods: Experiments and analysis. In D. R. Radev and E. H. Hovy, editors, Working Notes of the AAAI Spring Symposium on Intelligent Text Summarization, pages 60-68.",
	"links": null
	},
	"BIBREF37": {
	"ref_id": "b37",
	"title": "Cut and paste based summarization",
	"authors": [
	{
	"first": "Hongyan",
	"middle": [],
	"last": "Jing",
	"suffix": ""
	},
	{
	"first": "Kathleen",
	"middle": [
	"R"
	],
	"last": "Mckeown",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Proceedings of the Sixth Applied Natural Language Conference (ANLP-00) and the First Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL-00)",
	"volume": "",
	"issue": "",
	"pages": "178--185",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Jing, Hongyan and Kathleen R. McKeown. 2000. Cut and paste based summarization. In Proceedings of the Sixth Applied Natural Language Conference (ANLP-00) and the First Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL-00), pages 178-185.",
	"links": null
	},
	"BIBREF38": {
	"ref_id": "b38",
	"title": "Rhetoric of Everyday English Texts",
	"authors": [
	{
	"first": "Michael",
	"middle": [
	"P"
	],
	"last": "Jordan",
	"suffix": ""
	}
	],
	"year": 1984,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Jordan, Michael P. 1984. Rhetoric of Everyday English Texts. Allen and Unwin, London.",
	"links": null
	},
	"BIBREF39": {
	"ref_id": "b39",
	"title": "Role of verbs in document analysis",
	"authors": [
	{
	"first": "Judith",
	"middle": [
	"L"
	],
	"last": "Klavans",
	"suffix": ""
	},
	{
	"first": "Min-Yen",
	"middle": [],
	"last": "Kan",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and the 17th International Conference on Computational Linguistics (ACL/COLING-98)",
	"volume": "",
	"issue": "",
	"pages": "680--686",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Klavans, Judith L. and Min-Yen Kan. 1998. Role of verbs in document analysis. In Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and the 17th International Conference on Computational Linguistics (ACL/COLING-98), pages 680-686.",
	"links": null
	},
	"BIBREF40": {
	"ref_id": "b40",
	"title": "Statistics-based summarization-Step one: Sentence compression",
	"authors": [
	{
	"first": "Kevin",
	"middle": [],
	"last": "Knight",
	"suffix": ""
	},
	{
	"first": "Daniel",
	"middle": [],
	"last": "Marcu",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Proceedings of the 17th National Conference of the American Association for Artificial Intelligence (AAAI-2000)",
	"volume": "",
	"issue": "",
	"pages": "703--710",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Knight, Kevin and Daniel Marcu. 2000. Statistics-based summarization-Step one: Sentence compression. In Proceedings of the 17th National Conference of the American Association for Artificial Intelligence (AAAI-2000), pages 703-710.",
	"links": null
	},
	"BIBREF41": {
	"ref_id": "b41",
	"title": "Content Analysis: An Introduction to Its Methodology",
	"authors": [
	{
	"first": "Klaus",
	"middle": [],
	"last": "Krippendorff",
	"suffix": ""
	}
	],
	"year": 1980,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Krippendorff, Klaus. 1980. Content Analysis: An Introduction to Its Methodology. Sage Publications, Beverly Hills, CA.",
	"links": null
	},
	"BIBREF42": {
	"ref_id": "b42",
	"title": "A trainable document summarizer",
	"authors": [
	{
	"first": "Julian",
	"middle": [],
	"last": "Kupiec",
	"suffix": ""
	},
	{
	"first": "Jan",
	"middle": [
	"O"
	],
	"last": "Pedersen",
	"suffix": ""
	},
	{
	"first": "Francine",
	"middle": [],
	"last": "Chen",
	"suffix": ""
	}
	],
	"year": 1995,
	"venue": "Proceedings of the 18th Annual International Conference on Research and Development in Information Retrieval (SIGIR-95)",
	"volume": "",
	"issue": "",
	"pages": "68--73",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Kupiec, Julian, Jan O. Pedersen, and Francine Chen. 1995. A trainable document summarizer. In Proceedings of the 18th Annual International Conference on Research and Development in Information Retrieval (SIGIR-95), pages 68-73.",
	"links": null
	},
	"BIBREF43": {
	"ref_id": "b43",
	"title": "Indexing and Abstracting in Theory and Practice",
	"authors": [
	{
	"first": "Frederick",
	"middle": [],
	"last": "Lancaster",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Wilfrid",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Library Association",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Lancaster, Frederick Wilfrid. 1998. Indexing and Abstracting in Theory and Practice. Library Association, London.",
	"links": null
	},
	"BIBREF44": {
	"ref_id": "b44",
	"title": "The measurement of observer agreement for categorical data",
	"authors": [
	{
	"first": "J",
	"middle": [],
	"last": "Landis",
	"suffix": ""
	},
	{
	"first": "Gary",
	"middle": [
	"G"
	],
	"last": "Richard",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Koch",
	"suffix": ""
	}
	],
	"year": 1977,
	"venue": "Biometrics",
	"volume": "33",
	"issue": "",
	"pages": "159--174",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Landis, J. Richard and Gary G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33:159-174.",
	"links": null
	},
	"BIBREF45": {
	"ref_id": "b45",
	"title": "Digital libraries and autonomous citation indexing",
	"authors": [
	{
	"first": "Steve",
	"middle": [],
	"last": "Lawrence",
	"suffix": ""
	},
	{
	"first": "C",
	"middle": [
	"Lee"
	],
	"last": "Giles",
	"suffix": ""
	},
	{
	"first": "Kurt",
	"middle": [],
	"last": "Bollacker",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "IEEE Computer",
	"volume": "32",
	"issue": "6",
	"pages": "67--71",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Lawrence, Steve, C. Lee Giles, and Kurt Bollacker. 1999. Digital libraries and autonomous citation indexing. IEEE Computer, 32(6):67-71.",
	"links": null
	},
	"BIBREF46": {
	"ref_id": "b46",
	"title": "English Verb Classes and Alternations",
	"authors": [
	{
	"first": "Beth",
	"middle": [],
	"last": "Levin",
	"suffix": ""
	}
	],
	"year": 1991,
	"venue": "Speech and Natural Language: Proceedings of the ARPA Workshop of Human Language Technology",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Levin, Beth. 1993. English Verb Classes and Alternations. University of Chicago Press, Chicago. Lewis, David D. 1991. Evaluating text categorisation. In Speech and Natural Language: Proceedings of the ARPA Workshop of Human Language Technology.",
	"links": null
	},
	"BIBREF47": {
	"ref_id": "b47",
	"title": "The discourse-level structure of empirical abstracts: An exploratory study. Information Processing and Management",
	"authors": [
	{
	"first": "Elizabeth",
	"middle": [],
	"last": "Liddy",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Duross",
	"suffix": ""
	}
	],
	"year": 1991,
	"venue": "",
	"volume": "27",
	"issue": "",
	"pages": "55--81",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Liddy, Elizabeth DuRoss. 1991. The discourse-level structure of empirical abstracts: An exploratory study. Information Processing and Management, 27(1):55-81.",
	"links": null
	},
	"BIBREF48": {
	"ref_id": "b48",
	"title": "Identifying topics by position",
	"authors": [
	{
	"first": "Chin-Yew",
	"middle": [],
	"last": "Lin",
	"suffix": ""
	},
	{
	"first": "Eduard",
	"middle": [
	"H"
	],
	"last": "Hovy",
	"suffix": ""
	}
	],
	"year": 1997,
	"venue": "Proceedings of the Fifth Applied Natural Language Conference (ANLP-97)",
	"volume": "",
	"issue": "",
	"pages": "283--290",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Lin, Chin-Yew and Eduard H. Hovy. 1997. Identifying topics by position. In Proceedings of the Fifth Applied Natural Language Conference (ANLP-97), pages 283-290.",
	"links": null
	},
	"BIBREF49": {
	"ref_id": "b49",
	"title": "The automatic creation of literature abstracts",
	"authors": [
	{
	"first": "Hans",
	"middle": [],
	"last": "Luhn",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Peter",
	"suffix": ""
	}
	],
	"year": 1958,
	"venue": "IBM Journal of Research and Development",
	"volume": "2",
	"issue": "2",
	"pages": "159--165",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Luhn, Hans Peter. 1958. The automatic creation of literature abstracts. IBM Journal of Research and Development, 2(2):159-165.",
	"links": null
	},
	"BIBREF50": {
	"ref_id": "b50",
	"title": "The TIPSTER SUMMAC text summarization evaluation",
	"authors": [
	{
	"first": "Inderjeet",
	"middle": [],
	"last": "Mani",
	"suffix": ""
	},
	{
	"first": "Therese",
	"middle": [],
	"last": "Firmin",
	"suffix": ""
	},
	{
	"first": "David",
	"middle": [],
	"last": "House",
	"suffix": ""
	},
	{
	"first": "Gary",
	"middle": [],
	"last": "Klein",
	"suffix": ""
	},
	{
	"first": "Beth",
	"middle": [],
	"last": "Sundheim",
	"suffix": ""
	},
	{
	"first": "Lynette",
	"middle": [],
	"last": "Hirschman",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the Ninth Meeting of the European Chapter of the Association for Computational Linguistics (EACL-99)",
	"volume": "",
	"issue": "",
	"pages": "77--85",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Mani, Inderjeet, Therese Firmin, David House, Gary Klein, Beth Sundheim, and Lynette Hirschman. 1999. The TIPSTER SUMMAC text summarization evaluation. In Proceedings of the Ninth Meeting of the European Chapter of the Association for Computational Linguistics (EACL-99), pages 77-85.",
	"links": null
	},
	"BIBREF51": {
	"ref_id": "b51",
	"title": "Improving summaries by revising them",
	"authors": [
	{
	"first": "Inderjeet",
	"middle": [],
	"last": "Mani",
	"suffix": ""
	},
	{
	"first": "Barbara",
	"middle": [],
	"last": "Gates",
	"suffix": ""
	},
	{
	"first": "Eric",
	"middle": [],
	"last": "Bloedorn",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics (ACL-99)",
	"volume": "",
	"issue": "",
	"pages": "558--565",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Mani, Inderjeet, Barbara Gates, and Eric Bloedorn. 1999. Improving summaries by revising them. In Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics (ACL-99), pages 558-565.",
	"links": null
	},
	"BIBREF52": {
	"ref_id": "b52",
	"title": "Rhetorical structure theory: Description and construction of text structures",
	"authors": [
	{
	"first": "William",
	"middle": [
	"C"
	],
	"last": "Mann",
	"suffix": ""
	},
	{
	"first": "Sandra",
	"middle": [
	"A"
	],
	"last": "Thompson",
	"suffix": ""
	}
	],
	"year": 1987,
	"venue": "Natural Language Generation: New Results in Artificial Intelligence, Psychology, and Linguistics",
	"volume": "",
	"issue": "",
	"pages": "85--95",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Mann, William C. and Sandra A. Thompson. 1987. Rhetorical structure theory: Description and construction of text structures. In Gerard Kempen, editor, Natural Language Generation: New Results in Artificial Intelligence, Psychology, and Linguistics. Martinus Nijhoff Publishers, Dordrecht, the Netherlands, pages 85-95.",
	"links": null
	},
	"BIBREF53": {
	"ref_id": "b53",
	"title": "Discourse trees are good indicators of importance in text",
	"authors": [
	{
	"first": "Daniel",
	"middle": [],
	"last": "Marcu",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Advances in Automatic Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "123--136",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Marcu, Daniel. 1999. Discourse trees are good indicators of importance in text. In I. Mani and M. T. Maybury, editors, Advances in Automatic Text Summarization. MIT Press, Cambridge, pages 123-136.",
	"links": null
	},
	"BIBREF54": {
	"ref_id": "b54",
	"title": "Training algorithms for linear text classifiers",
	"authors": [
	{
	"first": "Andrew",
	"middle": [],
	"last": "Mccallum",
	"suffix": ""
	}
	],
	"year": 1997,
	"venue": "Proceedings of the 19th Annual International Conference on Research and Development in Information Retrieval (SIGIR-97)",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "McCallum, Andrew. 1997. Training algorithms for linear text classifiers. In Proceedings of the 19th Annual International Conference on Research and Development in Information Retrieval (SIGIR-97).",
	"links": null
	},
	"BIBREF55": {
	"ref_id": "b55",
	"title": "Relevance: The whole history",
	"authors": [
	{
	"first": "Stefano",
	"middle": [],
	"last": "Mizzaro",
	"suffix": ""
	}
	],
	"year": 1997,
	"venue": "Journal of the American Society for Information Science",
	"volume": "48",
	"issue": "9",
	"pages": "810--832",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Mizzaro, Stefano. 1997. Relevance: The whole history. Journal of the American Society for Information Science, 48(9):810-832.",
	"links": null
	},
	"BIBREF56": {
	"ref_id": "b56",
	"title": "Lexical cohesion computed by thesaural relations as an indicator of the structure of text",
	"authors": [
	{
	"first": "Jane",
	"middle": [],
	"last": "Morris",
	"suffix": ""
	},
	{
	"first": "Graeme",
	"middle": [],
	"last": "Hirst",
	"suffix": ""
	}
	],
	"year": 1991,
	"venue": "Computational Linguistics",
	"volume": "17",
	"issue": "1",
	"pages": "21--48",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Morris, Jane and Graeme Hirst. 1991. Lexical cohesion computed by thesaural relations as an indicator of the structure of text. Computational Linguistics, 17(1):21-48.",
	"links": null
	},
	"BIBREF57": {
	"ref_id": "b57",
	"title": "Speech acts and scientific facts",
	"authors": [
	{
	"first": "Greg",
	"middle": [],
	"last": "Myers",
	"suffix": ""
	}
	],
	"year": 1992,
	"venue": "Journal of Pragmatics",
	"volume": "17",
	"issue": "4",
	"pages": "295--313",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Myers, Greg. 1992. In this paper we report . . . -Speech acts and scientific facts. Journal of Pragmatics, 17(4):295-313.",
	"links": null
	},
	"BIBREF58": {
	"ref_id": "b58",
	"title": "Towards multi-paper summarization using reference information",
	"authors": [
	{
	"first": "Hidetsugu",
	"middle": [],
	"last": "Nanba",
	"suffix": ""
	},
	{
	"first": "Manabu",
	"middle": [],
	"last": "Okumura",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of IJCAI-99",
	"volume": "",
	"issue": "",
	"pages": "926--931",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Nanba, Hidetsugu and Manabu Okumura. 1999. Towards multi-paper summarization using reference information. In Proceedings of IJCAI-99, pages 926-931.",
	"links": null
	},
	"BIBREF59": {
	"ref_id": "b59",
	"title": "Constructing literature abstracts by computer: Techniques and prospects. Information Processing and Management",
	"authors": [
	{
	"first": "Chris",
	"middle": [
	"D"
	],
	"last": "Paice",
	"suffix": ""
	}
	],
	"year": 1990,
	"venue": "",
	"volume": "26",
	"issue": "",
	"pages": "171--186",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Paice, Chris D. 1990. Constructing literature abstracts by computer: Techniques and prospects. Information Processing and Management, 26(1):171-186.",
	"links": null
	},
	"BIBREF60": {
	"ref_id": "b60",
	"title": "The identification of important concepts in highly structured technical papers",
	"authors": [
	{
	"first": "Chris",
	"middle": [
	"D"
	],
	"last": "Paice",
	"suffix": ""
	},
	{
	"first": "A",
	"middle": [
	"Paul"
	],
	"last": "Jones",
	"suffix": ""
	}
	],
	"year": 1993,
	"venue": "Proceedings of the 16th Annual International Conference on Research and Development in Information Retrieval (SIGIR-93)",
	"volume": "",
	"issue": "",
	"pages": "69--78",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Paice, Chris D. and A. Paul Jones. 1993. The identification of important concepts in highly structured technical papers. In Proceedings of the 16th Annual International Conference on Research and Development in Information Retrieval (SIGIR-93), pages 69-78.",
	"links": null
	},
	"BIBREF61": {
	"ref_id": "b61",
	"title": "Automatic abstracting research at the chemical abstracts service",
	"authors": [
	{
	"first": "Joseph",
	"middle": [
	"J"
	],
	"last": "Pollock",
	"suffix": ""
	},
	{
	"first": "Antonio",
	"middle": [],
	"last": "Zamora",
	"suffix": ""
	}
	],
	"year": 1975,
	"venue": "Journal of Chemical Information and Computer Sciences",
	"volume": "15",
	"issue": "4",
	"pages": "226--232",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Pollock, Joseph J. and Antonio Zamora. 1975. Automatic abstracting research at the chemical abstracts service. Journal of Chemical Information and Computer Sciences, 15(4):226-232.",
	"links": null
	},
	"BIBREF62": {
	"ref_id": "b62",
	"title": "Generating natural language summaries from multiple on-line sources",
	"authors": [
	{
	"first": "Dragomir",
	"middle": [
	"R"
	],
	"last": "Radev",
	"suffix": ""
	},
	{
	"first": "R",
	"middle": [],
	"last": "Kathleen",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Mckeown",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Computational Linguistics",
	"volume": "24",
	"issue": "3",
	"pages": "469--500",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Radev, Dragomir R. and Kathleen R. McKeown. 1998. Generating natural language summaries from multiple on-line sources. Computational Linguistics, 24(3):469-500.",
	"links": null
	},
	"BIBREF63": {
	"ref_id": "b63",
	"title": "The formation of abstracts by the selection of sentences",
	"authors": [
	{
	"first": "G",
	"middle": [
	"J"
	],
	"last": "Rath",
	"suffix": ""
	},
	{
	"first": "A",
	"middle": [],
	"last": "Resnick",
	"suffix": ""
	},
	{
	"first": "T",
	"middle": [
	"R"
	],
	"last": "Savage",
	"suffix": ""
	}
	],
	"year": 1961,
	"venue": "American Documentation",
	"volume": "12",
	"issue": "2",
	"pages": "139--143",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Rath, G. J., A. Resnick, and T. R. Savage. 1961. The formation of abstracts by the selection of sentences. American Documentation, 12(2):139-143.",
	"links": null
	},
	"BIBREF64": {
	"ref_id": "b64",
	"title": "Measuring verb similarity",
	"authors": [
	{
	"first": "Philip",
	"middle": [],
	"last": "Resnik",
	"suffix": ""
	},
	{
	"first": "Mona",
	"middle": [],
	"last": "Diab",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Twenty-Second Annual Meeting of the Cognitive Science Society (COGSCI2000)",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Resnik, Philip and Mona Diab. 2000. Measuring verb similarity. In Twenty-Second Annual Meeting of the Cognitive Science Society (COGSCI2000).",
	"links": null
	},
	"BIBREF65": {
	"ref_id": "b65",
	"title": "Passive voice and rhetorical role in scientific writing",
	"authors": [
	{
	"first": "Kathryn",
	"middle": [],
	"last": "Riley",
	"suffix": ""
	}
	],
	"year": 1991,
	"venue": "Journal of Technical Writing and Communication",
	"volume": "21",
	"issue": "3",
	"pages": "239--257",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Riley, Kathryn. 1991. Passive voice and rhetorical role in scientific writing. Journal of Technical Writing and Communication, 21(3):239-257.",
	"links": null
	},
	"BIBREF66": {
	"ref_id": "b66",
	"title": "Abstracting and Indexing",
	"authors": [
	{
	"first": "Jennifer",
	"middle": [],
	"last": "Rowley",
	"suffix": ""
	}
	],
	"year": 1982,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Rowley, Jennifer. 1982. Abstracting and Indexing. Bingley, London.",
	"links": null
	},
	"BIBREF67": {
	"ref_id": "b67",
	"title": "Selective analysis for automatic abstracting: Evaluating indicativeness and acceptability",
	"authors": [
	{
	"first": "Horacio",
	"middle": [],
	"last": "Saggion",
	"suffix": ""
	},
	{
	"first": "Guy",
	"middle": [],
	"last": "Lapalme",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Proceedings of Content-Based Multimedia Information Access (RIAO)",
	"volume": "",
	"issue": "",
	"pages": "747--764",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Saggion, Horacio and Guy Lapalme. 2000. Selective analysis for automatic abstracting: Evaluating indicativeness and acceptability. In Proceedings of Content-Based Multimedia Information Access (RIAO), pages 747-764.",
	"links": null
	},
	"BIBREF68": {
	"ref_id": "b68",
	"title": "Introduction to Modern Information Retrieval",
	"authors": [
	{
	"first": "Gerard",
	"middle": [],
	"last": "Salton",
	"suffix": ""
	},
	{
	"first": "Michael",
	"middle": [
	"J"
	],
	"last": "Mcgill",
	"suffix": ""
	}
	],
	"year": 1983,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Salton, Gerard and Michael J. McGill. 1983. Introduction to Modern Information Retrieval. McGraw-Hill, Tokyo.",
	"links": null
	},
	"BIBREF69": {
	"ref_id": "b69",
	"title": "Automatically selecting useful phrases for dialogue act tagging",
	"authors": [
	{
	"first": "Ken",
	"middle": [],
	"last": "Samuel",
	"suffix": ""
	},
	{
	"first": "Sandra",
	"middle": [],
	"last": "Carberry",
	"suffix": ""
	},
	{
	"first": "K",
	"middle": [],
	"last": "Vijay-Shanker",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the Pacific Association for Computational Linguistics",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Samuel, Ken, Sandra Carberry, and K. Vijay-Shanker. 1999. Automatically selecting useful phrases for dialogue act tagging. In Proceedings of the Pacific Association for Computational Linguistics (PACLING-99).",
	"links": null
	},
	"BIBREF70": {
	"ref_id": "b70",
	"title": "Relevance: A review of and a framework for the thinking on the notion in information science",
	"authors": [
	{
	"first": "Tefko",
	"middle": [],
	"last": "Saracevic",
	"suffix": ""
	}
	],
	"year": 1975,
	"venue": "Journal of the American Society for Information Science",
	"volume": "26",
	"issue": "6",
	"pages": "321--343",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Saracevic, Tefko. 1975. Relevance: A review of and a framework for the thinking on the notion in information science. Journal of the American Society for Information Science, 26(6):321-343.",
	"links": null
	},
	"BIBREF71": {
	"ref_id": "b71",
	"title": "Nonparametric Statistics for the Behavioral Sciences",
	"authors": [
	{
	"first": "Sidney",
	"middle": [],
	"last": "Siegel",
	"suffix": ""
	},
	{
	"first": "N. John",
	"middle": [],
	"last": "Castellan",
	"suffix": ""
	}
	],
	"year": 1988,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Siegel, Sidney and N. John Castellan Jr. 1988. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, Berkeley, CA, second edition.",
	"links": null
	},
	"BIBREF72": {
	"ref_id": "b72",
	"title": "What sort of thing is an AI experiment",
	"authors": [
	{
	"first": "Sparck",
	"middle": [],
	"last": "Jones",
	"suffix": ""
	},
	{
	"first": "Karen",
	"middle": [],
	"last": "",
	"suffix": ""
	}
	],
	"year": 1990,
	"venue": "The Foundations of Artificial Intelligence: A SourceBook",
	"volume": "",
	"issue": "",
	"pages": "274--281",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Sparck Jones, Karen. 1990. What sort of thing is an AI experiment? In D. Partridge and Yorick Wilks, editors, The Foundations of Artificial Intelligence: A SourceBook. Cambridge University Press, Cambridge, pages 274-281.",
	"links": null
	},
	"BIBREF73": {
	"ref_id": "b73",
	"title": "Automatic summarising: Factors and directions",
	"authors": [
	{
	"first": "Sparck",
	"middle": [],
	"last": "Jones",
	"suffix": ""
	},
	{
	"first": "Karen",
	"middle": [],
	"last": "",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Advances in Automatic Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "1--12",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Sparck Jones, Karen. 1999. Automatic summarising: Factors and directions. In I. Mani and M. T. Maybury, editors, Advances in Automatic Text Summarization. MIT Press, Cambridge, pages 1-12.",
	"links": null
	},
	"BIBREF74": {
	"ref_id": "b74",
	"title": "Research articles in English",
	"authors": [
	{
	"first": "John",
	"middle": [],
	"last": "Swales",
	"suffix": ""
	}
	],
	"year": 1990,
	"venue": "Genre Analysis: English in Academic and Research Settings",
	"volume": "",
	"issue": "",
	"pages": "110--176",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Swales, John, 1990. Research articles in English. In Genre Analysis: English in Academic and Research Settings. Cambridge University Press, Cambridge, chapter 7, pages 110-176.",
	"links": null
	},
	"BIBREF75": {
	"ref_id": "b75",
	"title": "Argumentative Zoning: Information Extraction from Scientific Text",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Teufel, Simone, 1999. Argumentative Zoning: Information Extraction from Scientific Text. Ph.D. thesis, School of Cognitive Science, University of Edinburgh, Edinburgh.",
	"links": null
	},
	"BIBREF76": {
	"ref_id": "b76",
	"title": "Task-based evaluation of summary quality: Describing relationships between scientific papers",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": ""
	}
	],
	"year": 2001,
	"venue": "Proceedings of NAACL-01 Workshop \"Automatic Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Teufel, Simone. 2001. Task-based evaluation of summary quality: Describing relationships between scientific papers. In Proceedings of NAACL-01 Workshop \"Automatic Text Summarization.\"",
	"links": null
	},
	"BIBREF77": {
	"ref_id": "b77",
	"title": "An annotation scheme for discourse-level argumentation in research articles",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": ""
	},
	{
	"first": "Jean",
	"middle": [],
	"last": "Carletta",
	"suffix": ""
	},
	{
	"first": "Marc",
	"middle": [],
	"last": "Moens",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the Eighth Meeting of the European Chapter of the Association for Computational Linguistics (EACL-99)",
	"volume": "",
	"issue": "",
	"pages": "110--117",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Teufel, Simone, Jean Carletta, and Marc Moens. 1999. An annotation scheme for discourse-level argumentation in research articles. In Proceedings of the Eighth Meeting of the European Chapter of the Association for Computational Linguistics (EACL-99), pages 110-117.",
	"links": null
	},
	"BIBREF78": {
	"ref_id": "b78",
	"title": "Sentence extraction as a classification task",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": ""
	},
	{
	"first": "Marc",
	"middle": [],
	"last": "Moens",
	"suffix": ""
	}
	],
	"year": 1997,
	"venue": "Proceedings of the ACL/EACL-97 Workshop on Intelligent Scalable Text Summarization",
	"volume": "",
	"issue": "",
	"pages": "58--65",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Teufel, Simone and Marc Moens. 1997. Sentence extraction as a classification task. In Proceedings of the ACL/EACL-97 Workshop on Intelligent Scalable Text Summarization, pages 58-65.",
	"links": null
	},
	"BIBREF79": {
	"ref_id": "b79",
	"title": "What's yours and what's mine: Determining intellectual attribution in scientific text",
	"authors": [
	{
	"first": "Simone",
	"middle": [],
	"last": "Teufel",
	"suffix": ""
	},
	{
	"first": "Marc",
	"middle": [],
	"last": "Moens",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "Proceedings of the Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Teufel, Simone and Marc Moens. 2000. What's yours and what's mine: Determining intellectual attribution in scientific text. In Proceedings of the Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora.",
	"links": null
	},
	"BIBREF80": {
	"ref_id": "b80",
	"title": "Evaluation in the reporting verbs used in academic papers",
	"authors": [
	{
	"first": "Geoff",
	"middle": [],
	"last": "Thompson",
	"suffix": ""
	},
	{
	"first": "Ye",
	"middle": [],
	"last": "Yiyun",
	"suffix": ""
	}
	],
	"year": 1991,
	"venue": "Applied Linguistics",
	"volume": "12",
	"issue": "4",
	"pages": "365--382",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Thompson, Geoff and Ye Yiyun. 1991. Evaluation in the reporting verbs used in academic papers. Applied Linguistics, 12(4):365-382.",
	"links": null
	},
	"BIBREF81": {
	"ref_id": "b81",
	"title": "Advantages of query biased summaries",
	"authors": [
	{
	"first": "Anastasios",
	"middle": [],
	"last": "Tombros",
	"suffix": ""
	},
	{
	"first": "Mark",
	"middle": [],
	"last": "Sanderson",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Proceedings of the 21st",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Tombros, Anastasios, and Mark Sanderson. 1998. Advantages of query biased summaries. In Proceedings of the 21st",
	"links": null
	},
	"BIBREF82": {
	"ref_id": "b82",
	"title": "Annual International Conference on Research and Development in Information Retrieval (SIGIR-98). Association of Computing Machinery",
	"authors": [],
	"year": null,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Annual International Conference on Research and Development in Information Retrieval (SIGIR-98). Association of Computing Machinery.",
	"links": null
	},
	"BIBREF83": {
	"ref_id": "b83",
	"title": "Macrostructures: An Interdisciplinary Study of Global Structures in Discourse, Interaction and Cognition",
	"authors": [
	{
	"first": "Bogdan",
	"middle": [],
	"last": "Trawinski",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Van Dijk",
	"suffix": ""
	},
	{
	"first": "A",
	"middle": [],
	"last": "Teun",
	"suffix": ""
	}
	],
	"year": 1979,
	"venue": "Information Processing and Management",
	"volume": "25",
	"issue": "6",
	"pages": "693--702",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Trawinski, Bogdan. 1989. A methodology for writing problem-structured abstracts. Information Processing and Management, 25(6):693-702. van Dijk, Teun A. 1980. Macrostructures: An Interdisciplinary Study of Global Structures in Discourse, Interaction and Cognition. Lawrence Erlbaum, Hillsdale, NJ. van Rijsbergen, Cornelis Joost. 1979. Information Retrieval. Butterworth, London, second edition.",
	"links": null
	},
	"BIBREF84": {
	"ref_id": "b84",
	"title": "Tracking point of view in narrative",
	"authors": [
	{
	"first": "",
	"middle": [],
	"last": "Wiebe Janyce",
	"suffix": ""
	}
	],
	"year": 1994,
	"venue": "Computational Linguistics",
	"volume": "20",
	"issue": "2",
	"pages": "223--287",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Wiebe Janyce. 1994. Tracking point of view in narrative. Computational Linguistics, 20(2):223-287.",
	"links": null
	},
	"BIBREF85": {
	"ref_id": "b85",
	"title": "A re-examination of text categorization methods",
	"authors": [
	{
	"first": "Yiming",
	"middle": [],
	"last": "Yang",
	"suffix": ""
	},
	{
	"first": "Xin",
	"middle": [],
	"last": "Liu",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Proceedings of the 22nd Annual International Conference on Research and Development in Information Retrieval (SIGIR-99)",
	"volume": "",
	"issue": "",
	"pages": "42--49",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Yang, Yiming and Xin Liu. 1999. A re-examination of text categorization methods. In Proceedings of the 22nd Annual International Conference on Research and Development in Information Retrieval (SIGIR-99), pages 42-49.",
	"links": null
	},
	"BIBREF86": {
	"ref_id": "b86",
	"title": "A rhetoric for research in sciences and technologies",
	"authors": [
	{
	"first": "James",
	"middle": [
	"P"
	],
	"last": "Zappen",
	"suffix": ""
	}
	],
	"year": 1983,
	"venue": "New Essays in Technical and Scientific Communication Research Theory Practice",
	"volume": "",
	"issue": "",
	"pages": "123--138",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Zappen, James P. 1983. A rhetoric for research in sciences and technologies. In Paul V. Anderson, R. John Brockman, and Carolyn R. Miller, editors, New Essays in Technical and Scientific Communication Research Theory Practice. Baywood, Farmingdale, NY, pages 123-138.",
	"links": null
	},
	"BIBREF87": {
	"ref_id": "b87",
	"title": "Automatic text abstracting by selecting relevant passages",
	"authors": [
	{
	"first": "Klaus",
	"middle": [],
	"last": "Zechner",
	"suffix": ""
	}
	],
	"year": 1995,
	"venue": "",
	"volume": "224",
	"issue": "",
	"pages": "318--324",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Zechner, Klaus. 1995. Automatic text abstracting by selecting relevant passages. Master's thesis, Centre for Cognitive Science, University of Edinburgh, Edinburgh. Ziman, John M. 1969. Information, communication, knowledge. Nature, 224:318-324.",
	"links": null
	}
	},
	"ref_entries": {
	"FIGREF0": {
	"text": "Expert summary, contrastive links.",
	"num": null,
	"uris": null,
	"type_str": "figure"
	},
	"FIGREF1": {
	"text": "work of the authors)? Does this sentence refer to new, Does the sentence describe general to be explained or linguistic example sentences? that describes the specific aim Does this sentence contain material of the paper? explicit reference to the structure of the paper? Does this sentence make TEXTUAL OTHER or comparison of the own work to it? of other work, or a contrast Does it describe a negative aspect or support for the current paper? Does this sentence mention other work as basis of work by the authors (excluding previous",
	"num": null,
	"uris": null,
	"type_str": "figure"
	},
	"FIGREF2": {
	"text": "Decision tree for rhetorical annotation.",
	"num": null,
	"uris": null,
	"type_str": "figure"
	},
	"FIGREF3": {
	"text": "Distribution of rhetorical categories (entire document).",
	"num": null,
	"uris": null,
	"type_str": "figure"
	},
	"TABREF0": {
	"type_str": "table",
	"num": null,
	"text": "Annotation scheme for rhetorical status.",
	"content": "<table><tr><td>AIM</td><td>Specific research goal of the current paper</td></tr><tr><td>TEXTUAL</td><td>Statements about section structure</td></tr><tr><td>OWN</td><td>(Neutral) description of own work presented in current paper: Method-</td></tr><tr><td/><td>ology, results, discussion</td></tr><tr><td colspan=\"2\">BACKGROUND Generally accepted scientific background</td></tr><tr><td>CONTRAST</td><td>Statements of comparison with or contrast to other work; weaknesses of</td></tr><tr><td/><td>other work</td></tr><tr><td>BASIS</td><td>Statements of agreement with other work or continuation of other work</td></tr><tr><td>OTHER</td><td>(Neutral) description of other researchers' work</td></tr></table>",
	"html": null
	},
	"TABREF2": {
	"type_str": "table",
	"num": null,
	"text": "Confusion matrix between annotators B and C.",
	"content": "<table><tr><td/><td/><td/><td colspan=\"3\">ANNOTATOR B</td><td/><td/><td/></tr><tr><td/><td colspan=\"7\">AIM CTR TXT OWN BKG BAS OTH</td><td>Total</td></tr><tr><td>ANNOTATOR C AIM</td><td>35</td><td>2</td><td>1</td><td>19</td><td>3</td><td/><td>2</td><td>62</td></tr><tr><td>CTR</td><td/><td>86</td><td/><td>31</td><td>16</td><td/><td>23</td><td>156</td></tr><tr><td>TXT</td><td/><td/><td>31</td><td>7</td><td/><td/><td>1</td><td>3 9</td></tr><tr><td>OWN</td><td>10</td><td>62</td><td colspan=\"2\">5 2,298</td><td>25</td><td>3</td><td>84</td><td>2,487</td></tr><tr><td>BKG</td><td/><td>5</td><td/><td>1 3</td><td>115</td><td/><td>20</td><td>153</td></tr><tr><td>BAS</td><td>2</td><td/><td/><td>1 8</td><td>1</td><td>18</td><td>14</td><td>53</td></tr><tr><td>OTH</td><td>1</td><td>18</td><td>2</td><td>55</td><td>10</td><td>1</td><td>412</td><td>499</td></tr><tr><td>Total</td><td>48</td><td>173</td><td colspan=\"2\">39 2,441</td><td>170</td><td>22</td><td>556</td><td>3,449</td></tr></table>",
	"html": null
	},
	"TABREF3": {
	"type_str": "table",
	"num": null,
	"text": "Overview of feature pool.",
	"content": "<table><tr><td>Type</td><td>Name</td><td>Feature Description</td><td>Feature Values</td></tr><tr><td>Absolute</td><td>1. Loc</td><td>Position of sentence in relation to 10</td><td>A-J</td></tr><tr><td>location</td><td/><td>segments</td><td/></tr><tr><td>Explicit</td><td>2. Section</td><td>Relative and absolute position of</td><td>7 values</td></tr><tr><td>structure</td><td>Struct</td><td>sentence within section (e.g., first</td><td/></tr><tr><td/><td/><td>sentence in section or somewhere in</td><td/></tr><tr><td/><td/><td>second third)</td><td/></tr><tr><td/><td>3. Para</td><td>Relative position of sentence within a</td><td>Initial, Medial, Final</td></tr><tr><td/><td>Struct</td><td>paragraph</td><td/></tr><tr><td/><td colspan=\"2\">4. Headline Type of headline of current section</td><td>15 prototypical headlines</td></tr><tr><td/><td/><td/><td>or Non-prototypical</td></tr><tr><td>Sentence</td><td>5. Length</td><td>Is the sentence longer than a certain</td><td>Yes or No</td></tr><tr><td>length</td><td/><td>threshold, measured in words?</td><td/></tr><tr><td>Content</td><td>6. Title</td><td>Does the sentence contain words also</td><td>Yes or No</td></tr><tr><td>features</td><td/><td>occurring in the title or headlines?</td><td/></tr><tr><td/><td>7. TFIDF</td><td>Does the sentence contain \"significant</td><td>Yes or No</td></tr><tr><td/><td/><td>terms\" as determined by the TFIDF</td><td/></tr><tr><td/><td/><td>measure?</td><td/></tr><tr><td colspan=\"2\">Verb syntax 8. Voice</td><td>Voice (of first finite verb in sentence)</td><td>Active or Passive or</td></tr><tr><td/><td/><td/><td>NoVerb</td></tr><tr><td/><td>9. Tense</td><td>Tense (of first finite verb in sentence)</td><td>9 simple and complex</td></tr><tr><td/><td/><td/><td>tenses or NoVerb</td></tr><tr><td/><td>10. Modal</td><td>Is the first finite verb modified by</td><td>Modal or NoModal or</td></tr><tr><td/><td/><td>modal auxiliary?</td><td>NoVerb</td></tr><tr><td>Citations</td><td>11. Cit</td><td>Does the sentence contain a citation or</td><td>{Citation (self), Citation</td></tr><tr><td/><td/><td>the name of an author contained in the</td><td>(other), Author Name,</td></tr><tr><td/><td/><td>reference list? If it contains a citation,</td><td>or None} \u00d7 {Beginning,</td></tr><tr><td/><td/><td>is it a self-citation? Whereabouts in the</td><td>Middle, End}</td></tr><tr><td/><td/><td>sentence does the citation occur?</td><td/></tr><tr><td>History</td><td colspan=\"2\">12. History Most probable previous category</td><td>7 Target Categories +</td></tr><tr><td/><td/><td/><td>\"BEGIN\"</td></tr><tr><td>Meta-</td><td colspan=\"2\">13. Formulaic Type of formulaic expression occur-</td><td>18 Types of Formulaic</td></tr><tr><td>discourse</td><td/><td>ring in sentence</td><td>Expressions + 9 Agent</td></tr><tr><td/><td/><td/><td>Types or None</td></tr><tr><td/><td>14. Agent</td><td>Type of agent</td><td>9 Agent Types or None</td></tr><tr><td/><td colspan=\"2\">15. SegAgent Type of agent</td><td>9 Agent Types or None</td></tr><tr><td/><td>16. Action</td><td>Type of action, with or without</td><td/></tr><tr><td/><td/><td>negation</td><td/></tr></table>",
	"html": null
	},
	"TABREF5": {
	"type_str": "table",
	"num": null,
	"text": "Formulaic expression lexicon.",
	"content": "<table><tr><td>Indicator Type</td><td>Example</td><td>Number</td></tr><tr><td>GAP INTRODUCTION</td><td>to our knowledge</td><td>3</td></tr><tr><td>GENERAL FORMULAIC</td><td>in traditional approaches</td><td>10</td></tr><tr><td>DEIXIS</td><td>in this paper</td><td>11</td></tr><tr><td>SIMILARITY</td><td>similar to</td><td>56</td></tr><tr><td>COMPARISON</td><td>when compared to our</td><td>204</td></tr><tr><td>CONTRAST</td><td>however</td><td>6</td></tr><tr><td>DETAIL</td><td>this paper has also</td><td>4</td></tr><tr><td>METHOD</td><td>a novel method for VERB-ing</td><td>33</td></tr><tr><td>PREVIOUS CONTEXT</td><td>elsewhere, we have</td><td>25</td></tr><tr><td>FUTURE</td><td>avenue for improvement</td><td>16</td></tr><tr><td>AFFECT</td><td>hopefully</td><td>4</td></tr><tr><td>CONTINUATION</td><td>following the argument in</td><td>19</td></tr><tr><td>IN ORDER TO</td><td>in order to</td><td>1</td></tr><tr><td>POSITIVE ADJECTIVE</td><td>appealing</td><td>68</td></tr><tr><td>NEGATIVE ADJECTIVE</td><td>unsatisfactory</td><td>119</td></tr><tr><td>THEM FORMULAIC</td><td>along the lines of</td><td>6</td></tr><tr><td>TEXTSTRUCTURE</td><td>in section 3</td><td>16</td></tr><tr><td>NO TEXTSTRUCTURE</td><td>described in the last section</td><td>43</td></tr><tr><td>Total of 18 classes</td><td/><td>644</td></tr></table>",
	"html": null
	},
	"TABREF6": {
	"type_str": "table",
	"num": null,
	"text": "Agent lexicon.",
	"content": "<table><tr><td>Agent Type</td><td>Example</td><td colspan=\"2\">Number Removed</td></tr><tr><td>US AGENT</td><td>we</td><td>22</td><td/></tr><tr><td>THEM AGENT</td><td>his approach</td><td>21</td><td/></tr><tr><td>GENERAL AGENT</td><td>traditional methods</td><td>20</td><td>X</td></tr><tr><td>US PREVIOUS AGENT</td><td>the approach in SELFCITE</td><td>7</td><td/></tr><tr><td>OUR AIM AGENT</td><td>the point of this study</td><td>23</td><td/></tr><tr><td>REF US AGENT</td><td>this method (this WORK NOUN)</td><td>6</td><td/></tr><tr><td>REF AGENT</td><td>the paper</td><td>11</td><td>X</td></tr><tr><td colspan=\"2\">THEM PRONOUN AGENT they</td><td>1</td><td>X</td></tr><tr><td>AIM REF AGENT</td><td>its goal</td><td>8</td><td/></tr><tr><td>GAP AGENT</td><td>none of these papers</td><td>8</td><td/></tr><tr><td>PROBLEM AGENT</td><td>these drawbacks</td><td>3</td><td>X</td></tr><tr><td>SOLUTION AGENT</td><td>a way out of this dilemma</td><td>4</td><td>X</td></tr><tr><td colspan=\"2\">TEXTSTRUCTURE AGENT the concluding chapter</td><td>33</td><td/></tr><tr><td>Total of 13 classes</td><td/><td>167</td><td/></tr></table>",
	"html": null
	},
	"TABREF7": {
	"type_str": "table",
	"num": null,
	"text": "Action lexicon.",
	"content": "<table><tr><td>Action Type</td><td>Example</td><td colspan=\"2\">Number Removed</td></tr><tr><td>AFFECT</td><td>we hope to improve our results</td><td>9</td><td>X</td></tr><tr><td>ARGUMENTATION</td><td>we argue against a model of</td><td>19</td><td>X</td></tr><tr><td>AWARENESS</td><td>we are not aware of attempts</td><td>5</td><td>+</td></tr><tr><td>BETTER SOLUTION</td><td>our system outperforms . . .</td><td>9</td><td>\u2212</td></tr><tr><td>CHANGE</td><td>we extend CITE's algorithm</td><td>23</td><td/></tr><tr><td>COMPARISON</td><td>we tested our system against . . .</td><td>4</td><td/></tr><tr><td>CONTINUATION</td><td>we follow CITE . . .</td><td>13</td><td/></tr><tr><td>CONTRAST</td><td>our approach differs from . . .</td><td>12</td><td>\u2212</td></tr><tr><td>FUTURE INTEREST</td><td>we intend to improve . . .</td><td>4</td><td>X</td></tr><tr><td>INTEREST</td><td>we are concerned with . . .</td><td>28</td><td/></tr><tr><td>NEED</td><td>this approach, however, lacks . . .</td><td>8</td><td>X</td></tr><tr><td>PRESENTATION</td><td>we present here a method for . . .</td><td>19</td><td>\u2212</td></tr><tr><td>PROBLEM</td><td>this approach fails . . .</td><td>61</td><td>\u2212</td></tr><tr><td>RESEARCH</td><td>we collected our data from . . .</td><td>54</td><td/></tr><tr><td>SIMILAR</td><td>our approach resembles that of</td><td>13</td><td/></tr><tr><td>SOLUTION</td><td>we solve this problem by . . .</td><td>64</td><td/></tr><tr><td>TEXTSTRUCTURE</td><td>the paper is organized . . .</td><td>13</td><td/></tr><tr><td>USE</td><td>we employ CITE's method . . .</td><td>5</td><td/></tr><tr><td>COPULA</td><td>our goal is to . . .</td><td>1</td><td/></tr><tr><td>POSSESSION</td><td>we have three goals . . .</td><td>1</td><td/></tr><tr><td>Total of 20 classes</td><td/><td>365</td><td/></tr></table>",
	"html": null
	},
	"TABREF8": {
	"type_str": "table",
	"num": null,
	"text": "Performance per category: F-measure (F), precision (P) and recall (R).",
	"content": "<table><tr><td/><td>AIM</td><td colspan=\"2\">CONTR. TEXTUAL</td><td>OWN</td><td>BACKG.</td><td>BASIS</td><td>OTHER</td></tr><tr><td/><td>F P R</td><td>F P R F P</td><td colspan=\"4\">R F P R F P R F P R F P R</td></tr><tr><td>System</td><td colspan=\"6\">52 44 65 26 34 20 61 57 66 86 84 88 45 40 50 38 37 40 44 52 39</td></tr><tr><td colspan=\"7\">Baseline 11 30 7 17 31 12 23 56 15 83 78 90 22 32 17 7 15 5 44 47 42</td></tr><tr><td colspan=\"7\">Humans 63 72 56 52 50 55 79 79 79 93 94 92 71 68 75 48 82 34 78 74 83</td></tr></table>",
	"html": null
	},
	"TABREF9": {
	"type_str": "table",
	"num": null,
	"text": "Confusion matrix: Human versus automatic annotation.",
	"content": "<table><tr><td/><td/><td/><td/><td/><td colspan=\"2\">MACHINE</td><td/><td/><td/></tr><tr><td/><td/><td colspan=\"7\">AIM CTR TXT OWN BKG BAS OTH</td><td>Total</td></tr><tr><td>HUMAN</td><td>AIM</td><td>127</td><td>6</td><td>13</td><td>23</td><td>19</td><td>5</td><td>10</td><td>203</td></tr><tr><td/><td>CTR</td><td>21</td><td>112</td><td>4</td><td>204</td><td>87</td><td>18</td><td>126</td><td>572</td></tr><tr><td/><td>TXT</td><td>14</td><td>1</td><td>145</td><td>46</td><td>6</td><td>2</td><td>6</td><td>220</td></tr><tr><td/><td>OWN</td><td>100</td><td>108</td><td colspan=\"2\">84 7,231</td><td>222</td><td>71</td><td>424</td><td>8,240</td></tr><tr><td/><td>BKG</td><td>14</td><td>31</td><td>1</td><td>222</td><td>370</td><td>5</td><td>101</td><td>744</td></tr><tr><td/><td>BAS</td><td>17</td><td>7</td><td>7</td><td>60</td><td>8</td><td>97</td><td>39</td><td>235</td></tr><tr><td/><td>OTH</td><td>6</td><td>70</td><td>10</td><td>828</td><td>215</td><td>72</td><td>773</td><td>1,974</td></tr><tr><td/><td>Total</td><td>299</td><td>335</td><td colspan=\"2\">264 8,614</td><td>927</td><td colspan=\"2\">270 1,479</td><td>12,188</td></tr></table>",
	"html": null
	},
	"TABREF10": {
	"type_str": "table",
	"num": null,
	"text": "Overall classification results.",
	"content": "<table><tr><td/><td colspan=\"5\">System/Baseline Compared with One Human Annotator</td><td>3 Humans</td></tr><tr><td/><td colspan=\"5\">System Text Class. Random Random (Distr.) Most Freq.</td><td/></tr><tr><td>Kappa</td><td>.45</td><td>.30</td><td>\u2212.10</td><td>0</td><td>\u2212.13</td><td>.71</td></tr><tr><td>Accuracy</td><td>.73</td><td>.72</td><td>.14</td><td>.48</td><td>.67</td><td>.87</td></tr><tr><td>Macro-F</td><td>.50</td><td>.30</td><td>.09</td><td>.14</td><td>.11</td><td>.69</td></tr></table>",
	"html": null
	},
	"TABREF11": {
	"type_str": "table",
	"num": null,
	"text": "Precision and recall of rhetorical classification, individual features. In Hindle's proposal, words are similar if we have strong statistical evidence that they tend to participate in the same events.",
	"content": "<table><tr><td>Features</td><td/><td colspan=\"5\">Precision/Recall per Category (in %)</td><td/></tr><tr><td/><td colspan=\"2\">AIM CONTR.</td><td>TXT.</td><td colspan=\"4\">OWN BACKG. BASIS OTHER</td></tr><tr><td>SegAgent alone</td><td>-</td><td>17/0</td><td colspan=\"2\">-74/94</td><td>53/16</td><td>-</td><td>46/33</td></tr><tr><td>Agent alone</td><td>-</td><td>-</td><td colspan=\"2\">-71/93</td><td>-</td><td>-</td><td>36/23</td></tr><tr><td>Location alone</td><td>-</td><td>-</td><td colspan=\"2\">-74/97</td><td>40/36</td><td>-</td><td>28/9</td></tr><tr><td>Headlines alone</td><td>-</td><td>-</td><td colspan=\"2\">-75/95</td><td>-</td><td>-</td><td>29/25</td></tr><tr><td>Citation alone</td><td>-</td><td>-</td><td colspan=\"2\">-73/96</td><td>-</td><td>-</td><td>43/30</td></tr><tr><td colspan=\"2\">Formulaic alone 40/2</td><td colspan=\"3\">45/2 75/39 71/98</td><td>-</td><td>40/1</td><td>47/13</td></tr><tr><td>Action alone</td><td>-</td><td>43/1</td><td colspan=\"2\">-68/99</td><td>-</td><td>-</td><td>-</td></tr><tr><td>History alone</td><td>-</td><td>-</td><td>--</td><td>70/8</td><td>16/99</td><td>-</td><td>-</td></tr></table>",
	"html": null
	}
	}
	}
	}