| { |
| "paper_id": "D10-1031", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T15:51:34.694593Z" |
| }, |
| "title": "Automatic Discovery of Manner Relations and its Applications", |
| "authors": [ |
| { |
| "first": "Eduardo", |
| "middle": [], |
| "last": "Blanco", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "The University of Texas at Dallas Richardson", |
| "location": { |
| "postCode": "75080", |
| "region": "TX", |
| "country": "USA" |
| } |
| }, |
| "email": "eduardo@hlt.utdallas.edu" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Moldovan", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "The University of Texas at Dallas Richardson", |
| "location": { |
| "postCode": "75080", |
| "region": "TX", |
| "country": "USA" |
| } |
| }, |
| "email": "moldovan@hlt.utdallas.edu" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "This paper presents a method for the automatic discovery of MANNER relations from text. An extended definition of MANNER is proposed, including restrictions on the sorts of concepts that can be part of its domain and range. The connections with other relations and the lexico-syntactic patterns that encode MANNER are analyzed. A new feature set specialized on MANNER detection is depicted and justified. Experimental results show improvement over previous attempts to extract MANNER. Combinations of MANNER with other semantic relations are also discussed.", |
| "pdf_parse": { |
| "paper_id": "D10-1031", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "This paper presents a method for the automatic discovery of MANNER relations from text. An extended definition of MANNER is proposed, including restrictions on the sorts of concepts that can be part of its domain and range. The connections with other relations and the lexico-syntactic patterns that encode MANNER are analyzed. A new feature set specialized on MANNER detection is depicted and justified. Experimental results show improvement over previous attempts to extract MANNER. Combinations of MANNER with other semantic relations are also discussed.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "Extracting semantic relations from text is an important step towards understanding the meaning of text. Many applications that use no semantics, or only shallow semantics, could benefit by having available more text semantics. Recently, there is a growing interest in text semantics (M\u00e0rquez et al., 2008; Davidov and Rappoport, 2008) .", |
| "cite_spans": [ |
| { |
| "start": 283, |
| "end": 305, |
| "text": "(M\u00e0rquez et al., 2008;", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 306, |
| "end": 334, |
| "text": "Davidov and Rappoport, 2008)", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "An important semantic relation for many applications is the MANNER relation. Broadly speaking, MANNER encodes the mode, style, way or fashion in which something is done or happened. For example, quick delivery encodes a MANNER relation, since quick is the manner in which the delivery happened.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "An application of MANNER detection is Question Answering, where many how questions refer to this particular relation. Consider for example the question How did the President communicate his mes-sage?, and the text Through his spokesman, Obama sent a strong message [. . . ] . To answer such questions, it is useful to identify first the MANNER relations in text.", |
| "cite_spans": [ |
| { |
| "start": 265, |
| "end": 273, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "MANNER occurs frequently in text and it is expressed by a wide variety of lexico-syntactic patterns.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "For example, PropBank annotates 8,037 ARGM-MNR relations (10.7%) out of 74,980 adjunct-like arguments (ARGMs). There are verbs that state a particular way of doing something, e.g., to limp implicitly states a particular way of walking. Adverbial phrases and prepositional phrases are the most productive patterns, e.g., The nation's industrial sector is now growing very slowly if at all and He started the company on his own. Consider the following example: The company said Mr. Stronach will personally direct the restructuring assisted by Manfred Gingl, [. . . ] 1 . There are two MANNER relations in this sentence: the underlined chunks of text encode the way in which Mr. Stronach will direct the restructuring.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "The extraction of semantic relations in general has caught the attention of several researchers. Approaches to detect semantic relations usually focus on particular lexical and syntactic patterns. There are both unsupervised (Davidov et al., 2007; Turney, 2006) and supervised approaches. The SemEval-2007 Task 04 (Girju et al., 2007 aimed at relations between nominals. Work has been done on detecting relations within noun phrases (Nulty, 2007) , named entities (Hirano et al., 2007) , clauses (Szpakowicz et al., 1995) and syntax-based comma resolution (Srikumar et al., 2008) . There have been proposals to detect a particular relation, e.g., CAUSE (Chang and Choi, 2006) , INTENT (Tatu, 2005) , PART-WHOLE (Girju et al., 2006) and IS-A (Hearst, 1992) .", |
| "cite_spans": [ |
| { |
| "start": 225, |
| "end": 247, |
| "text": "(Davidov et al., 2007;", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 248, |
| "end": 261, |
| "text": "Turney, 2006)", |
| "ref_id": "BIBREF22" |
| }, |
| { |
| "start": 293, |
| "end": 305, |
| "text": "SemEval-2007", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 306, |
| "end": 333, |
| "text": "Task 04 (Girju et al., 2007", |
| "ref_id": null |
| }, |
| { |
| "start": 433, |
| "end": 446, |
| "text": "(Nulty, 2007)", |
| "ref_id": "BIBREF17" |
| }, |
| { |
| "start": 464, |
| "end": 485, |
| "text": "(Hirano et al., 2007)", |
| "ref_id": "BIBREF14" |
| }, |
| { |
| "start": 496, |
| "end": 521, |
| "text": "(Szpakowicz et al., 1995)", |
| "ref_id": "BIBREF20" |
| }, |
| { |
| "start": 556, |
| "end": 579, |
| "text": "(Srikumar et al., 2008)", |
| "ref_id": "BIBREF19" |
| }, |
| { |
| "start": 653, |
| "end": 675, |
| "text": "(Chang and Choi, 2006)", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 678, |
| "end": 684, |
| "text": "INTENT", |
| "ref_id": null |
| }, |
| { |
| "start": 685, |
| "end": 697, |
| "text": "(Tatu, 2005)", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 711, |
| "end": 731, |
| "text": "(Girju et al., 2006)", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 741, |
| "end": 755, |
| "text": "(Hearst, 1992)", |
| "ref_id": "BIBREF12" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Previous Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "MANNER is a frequent relation, but besides theoretical studies there is not much work on detecting it. Girju et al. (2003) propose a set of features to extract MANNER exclusively from adverbial phrases and report a precision of 64.44% and recall of 68.67%. MANNER is a semantic role, and all the works on the extraction of roles (Gildea and Jurafsky, 2002; Giuglea and Moschitti, 2006) extracts MANNER as well. However, these approaches treat MANNER as any other role and do not use any specific features for its detection. As we show in this paper, MANNER has its own unique characteristics and identifying them improves the extraction accuracy. The two most used semantic role annotation resources, FrameNet (Baker et al., 1998) and Prop-Bank (Palmer et al., 2005) , include MANNER.", |
| "cite_spans": [ |
| { |
| "start": 103, |
| "end": 122, |
| "text": "Girju et al. (2003)", |
| "ref_id": "BIBREF7" |
| }, |
| { |
| "start": 329, |
| "end": 356, |
| "text": "(Gildea and Jurafsky, 2002;", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 357, |
| "end": 385, |
| "text": "Giuglea and Moschitti, 2006)", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 710, |
| "end": 730, |
| "text": "(Baker et al., 1998)", |
| "ref_id": "BIBREF0" |
| }, |
| { |
| "start": 745, |
| "end": 766, |
| "text": "(Palmer et al., 2005)", |
| "ref_id": "BIBREF18" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Previous Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "The main contributions of this paper are: (1) empirical study of MANNER and its semantics;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Previous Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "(2) analysis of the differences with other relations;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Previous Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "(3) lexico-syntactic patterns expressing MANNER; (4) a set of features specialized on the detection of MANNER; and (5) the way MANNER combines with other semantic relations.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Previous Work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Traditionally, a semantic relation is defined by stating the kind of connection linking two concepts. For example, MANNER is loosely defined by the PropBank annotation guidelines 2 as manner adverbs specify how an action is performed [. . . ] manner should be used when an adverb be an answer to a question starting with 'how?'. We find this kind of definition weak and prone to confusion (Section 3.2). Nonetheless, to the best of our knowledge, semantic relations have been mostly defined stating only a vague definition.", |
| "cite_spans": [ |
| { |
| "start": 234, |
| "end": 242, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Semantics of MANNER Relation", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Following (Helbig, 2005) , we propose an extended definition for semantic relations, includ-ing semantic restrictions for its domain and range. These restrictions help deciding which relation holds between a given pair of concepts. A relation shall not hold between two concepts unless they belong to its domain and range. These restrictions are based on theoretical and empirical grounds.", |
| "cite_spans": [ |
| { |
| "start": 10, |
| "end": 24, |
| "text": "(Helbig, 2005)", |
| "ref_id": "BIBREF13" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "The Semantics of MANNER Relation", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Formally, MANNER is represented as MNR(x, y), and it should be read x is the manner in which y happened. In addition, DOMAIN(MNR) and RANGE(MNR) are the sets of sorts of concepts that can be part of the first and second argument. RANGE(MNR), namely y, is restricted to situations, which are defined as anything that happens at a time and place. Situations include events and states and can be expressed by verbs or nouns, e.g., conference, race, mix and grow. DOMAIN(MNR), namely x, is restricted to qualities (ql), non temporal abstract objects (ntao) and states (st). Qualities represent characteristics that can be assigned to other concepts, such as slowly and abruptly. Non temporal abstract objects are intangible entities. They are somehow product of human reasoning and are not palpable. They do not encode periods or points of time, such as week, or yesterday. For example, odor, disease, and mile are ntao; book and car are not because they are tangible. Unlike events, states are situations that do not imply a change in the concepts involved. For example, standing there or holding hands are states; whereas walking to the park and pinching him are events. For more details about these semantic classes, refer to (Helbig, 2005) .", |
| "cite_spans": [ |
| { |
| "start": 1225, |
| "end": 1239, |
| "text": "(Helbig, 2005)", |
| "ref_id": "BIBREF13" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "MANNER Definition", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "These semantic restrictions on MANNER come after studying previous definitions and manual examination of hundreds of examples. Their use and benefits are described in Section 4. Asking how does not identify MANNER in many cases. For example, given John broke the window [with a hammer], the question how did John break the window? can be answered by with the hammer, and yet the hammer is not the MANNER but the IN A way of solving this ambiguity is by prioritizing the semantic relations among the possible candidates for a given pair of concepts. For example, if both INSTRUMENT and MANNER are possible, the former is extracted. In a similar fashion, AT-LOCATION and AT-TIME could have higher priority than MANNER. This idea has one big disadvantage: the correct detection of MANNER relies on the detection of several other relations, a problem which has proven difficult and thus would unnecessarily introduce errors.", |
| "cite_spans": [ |
| { |
| "start": 412, |
| "end": 414, |
| "text": "IN", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "MANNER Definition", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "Using the proposed extended definition one may discard the false MANNER relations above. Hammer is not a quality, non temporal abstract object or state (hammers are palpable objects), so by definition a relation of the form MNR(the hammer, y) shall not hold. Similarly, fence and week do not fulfill the domain restriction, so MNR(over the fence, y) and MNR(every other week, y) are not valid either.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "MANNER and Other Semantic Relations", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "MANNER also relates to CAUSE. Again, asking how? does not resolve the ambiguity. Given The legislation itself noted that it [was introduced] x \"by request,\" [. . . ] (wsj 0041, 47), we believe the underlined PP indicates the CAUSE and not the MANNER of x because the introduction of the legislation is the effect of the request. Using the extended definition, since request is an event (it implies a change), MNR(by request, y) is discarded based on the domain and range restrictions.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "MANNER and Other Semantic Relations", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "In order to implement domain and range restrictions, one needs to map words to the four proposed semantic classes: situations (si), states (st), qualities (ql) and non temporal abstract objects (ntao). These classes are the ones involved in MNR; work has been done to define in a similar way more relations, but we do not report on that in this paper.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Argument Extraction", |
| "sec_num": "4" |
| }, |
| { |
| "text": "First, the head word of a potential argument is identified. Then, the head is mapped into a semantic class using three sources of information: POS tags, WordNet hypernyms and named entity (NE) types. Table 1 presents the rules that define the mapping. We obtained them following a data-driven ap-proach using a subset of MANNER annotation from PropBank and FrameNet. Intermediate classes are defined to facilitate legibility; intermediate classes ending in -NE only involve named entity types.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 200, |
| "end": 207, |
| "text": "Table 1", |
| "ref_id": "TABREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Argument Extraction", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Words are automatically POS tagged using a modified Brill tagger. We do not perform word sense disambiguation because in our experiments it did not bring any improvement; all senses are considered for each word. isHypo(x) for a given word w indicates if any of the senses of w is a hyponym of x in WordNet 2.0. An in-house NE recognizer is used to assign NE types. It detects 90 types organized in a hierarchy with an accuracy of 92% and it has been used in a state-of-the-art Question Answering system (Moldovan et al., 2007) . As far as the mapping is concerned, only the following NE types are used: human, organization, country, town, province, other-loc, money, date and time. The mapping also uses an automatically built list of verbs and nouns that encode events (verb events and noun events).", |
| "cite_spans": [ |
| { |
| "start": 503, |
| "end": 526, |
| "text": "(Moldovan et al., 2007)", |
| "ref_id": "BIBREF16" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Argument Extraction", |
| "sec_num": "4" |
| }, |
| { |
| "text": "The procedure to map words into semantic classes has been evaluated on a subset of PropBank which was not used to define the mapping. First, we selected 1,091 sentences which contained a total of 171 MANNER relations. We syntactically parsed the sentences using Charniak's parser and then performed argument detection by matching the trees to the syntactic patterns depicted in Section 5. 52,612 arguments pairs were detected as potential MAN-NER. Because of parsing errors, 146 (85.4%) of the 171 MANNER relations are in this set.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Argument Extraction", |
| "sec_num": "4" |
| }, |
| { |
| "text": "After mapping and enforcing domain and range constraints, the argument pairs were reduced to 11,724 (22.3%). The filtered subset includes 140 (81.8%) of the 171 MANNER relations. The filtering does make mistakes, but the massive pruning mainly filters out potential relations that do not hold: it filters 77.7% of argument pairs and it only misclassifies 6 pairs.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Argument Extraction", |
| "sec_num": "4" |
| }, |
| { |
| "text": "MANNER is expressed by a wide variety of lexicosyntactic patterns, implicitly or explicitly. Table 2 shows the syntactic distribution of MAN-NER relation in PropBank. We only consider relations between a single node in the syntactic tree and For both ADVP and PP the most common direct ancestor is either a VP or S, although examples are found that do not follow this rule. Table 3 shows the number of occurrences for several parent nodes and examples. Only taking into account phrases whose ancestor is either a VP or S yields a coverage of 98% and thus those are the focus of this work.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 93, |
| "end": 100, |
| "text": "Table 2", |
| "ref_id": "TABREF3" |
| }, |
| { |
| "start": 374, |
| "end": 381, |
| "text": "Table 3", |
| "ref_id": "TABREF5" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Lexico-Syntactic Patterns Expressing MANNER", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Both ADVPs and PPs are highly ambiguous when the task is to identify their semantics. The PropBank authors (Palmer et al., 2005) report discrepancies between annotators mainly with AT-LOCATION and simply no relation, i.e., when a phrase does not encode a role at all. In their annotation, 22.2% of AD-VPs encode MANNER (30.3% AT-TIME), whereas only 4.6% of PPs starting with in and 6.1% start- ing with at encode MANNER. The vast majority of PPs encode either a AT-TIME or AT-LOCATION. MANNER relations expressed by ADVPs are easier to detect since the adverb is a clear signal. Adverbs ending in -ly are more likely to encode a MAN-NER. Not surprisingly, the verb they attach to also plays an important role. Section 6.2 depicts the features used.", |
| "cite_spans": [ |
| { |
| "start": 107, |
| "end": 128, |
| "text": "(Palmer et al., 2005)", |
| "ref_id": "BIBREF18" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Ambiguities of MANNER", |
| "sec_num": "5.1" |
| }, |
| { |
| "text": "PPs are more complicated since the preposition per se is not a signal of whether or not the phrase encodes a MANNER. Even prepositions such as under and over can introduce a MANNER. For example, A majority of an NIH-appointed panel recommended late last year that the research continue under carefully controlled conditions, [. . . ] (wsj 0047, 9) and [. . . ] bars where Japanese revelers sing over recorded music, [. . . ] (wsj 0300, 3). Note that in both cases, the head of the NP contained in the PP encoding MANNER (conditions and music) belongs to ntao (Section 4). Other prepositions, like with and like are more likely to encode a MANNER, but again it is not guaranteed.", |
| "cite_spans": [ |
| { |
| "start": 325, |
| "end": 333, |
| "text": "[. . . ]", |
| "ref_id": null |
| }, |
| { |
| "start": 416, |
| "end": 424, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Ambiguities of MANNER", |
| "sec_num": "5.1" |
| }, |
| { |
| "text": "We propose a supervised learning approach, where instances are positive and negative MANNER examples. Due to their intrinsic difference, we build dif-ferent models for ADVPs and PPs.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "6" |
| }, |
| { |
| "text": "The corpus building procedure is as follows. First, all ADVPs and PPs whose parent node is a VP or S and encode a MANNER according to PropBank are extracted, yielding 3559 and 3499 positive instances respectively. Then, 10,000 examples of AD-VPs and another 10,000 of PPs from the Penn Tree-Bank not encoding a MANNER according to Prop-Bank are added. These negative instances must have as their parent node either VP or S as well and are selected randomly.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Building the Corpus", |
| "sec_num": "6.1" |
| }, |
| { |
| "text": "The total number of instances, 13,559 for ADVPs and 13,499 for PPs, are then divided into training (60%), held-out (20%) and test (20%). The held-out portion is used to tune the feature set and the final results provided are the ones obtained with the test portion, i.e., instances that have not been used in any way to learn the models. Because PropBank adds semantic role annotation on top of the Penn TreeBank, we have gold syntactic annotation for all instances.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Building the Corpus", |
| "sec_num": "6.1" |
| }, |
| { |
| "text": "Selected features are derived from previous works on detecting semantic roles, namely (Gildea and Jurafsky, 2002) CoNLL-2005 Shared Task (Carreras and M\u00e0rquez, 2005) , combined with new, manner-specific features that we introduce. These new features bring a significant improvement and are dependent on the phrase potentially encoding a MANNER. Experimentation has shown that MANNER relations expressed by an ADVP are easier to detect than the ones expressed by a PP.", |
| "cite_spans": [ |
| { |
| "start": 86, |
| "end": 113, |
| "text": "(Gildea and Jurafsky, 2002)", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 137, |
| "end": 165, |
| "text": "(Carreras and M\u00e0rquez, 2005)", |
| "ref_id": "BIBREF2" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Selecting features", |
| "sec_num": "6.2" |
| }, |
| { |
| "text": "The feature set used is depicted in Table 4 . Some features are typical of semantic role labeling, but features adverb, dictionary and ends-with-ly are specialized to MANNER extraction from ADVPs. These three additional features bring a significant improvement (Section 7).", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 36, |
| "end": 43, |
| "text": "Table 4", |
| "ref_id": "TABREF7" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "We only provide details for the non-obvious features.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "The main adverb and verb are retrieved by selecting the last adverb or verb of a sequence. For example, in more strongly, the main adverb is strongly, and in had been rescued the main verb is rescued.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Dictionary tests the presence of the adverb in a custom built dictionary which contains all lemmas for adverbs in WordNet whose gloss matches the regular expression in a .* (manner|way|fashion|style).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "For example, more.adv.1: used to form the comparative of some adjectives and adverbs does not belong to the dictionary, and strongly.adv.1: with strength or in a strong manner does. This feature is an extension of the dictionary presented in (Girju et al., 2003) .", |
| "cite_spans": [ |
| { |
| "start": 242, |
| "end": 262, |
| "text": "(Girju et al., 2003)", |
| "ref_id": "BIBREF7" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Given the sentence [. . . ] We [work [damn hard] ADVP at what we do for damn little pay] VP , and [. . . ] (wsj 1144, 128) , the features are: {parent-node:VP, num-leaves:2, adverb:hard, dictionary:no, ends-with-ly:no, POS-tag-bef:RB, POS-tag-aft:IN, verb:work, distance:1}, and it is a positive instance.", |
| "cite_spans": [ |
| { |
| "start": 19, |
| "end": 27, |
| "text": "[. . . ]", |
| "ref_id": null |
| }, |
| { |
| "start": 98, |
| "end": 122, |
| "text": "[. . . ] (wsj 1144, 128)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Prepositional Phrases PPs are known to be highly ambiguous and more features need to be added. The complete set is depicted in Table 5 .", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 127, |
| "end": 134, |
| "text": "Table 5", |
| "ref_id": "TABREF8" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Some features are typical of semantic role detection; we only provide a justification for the new features added. Num-pp-bef and num-pp-aft captures the number of PP siblings before and after the PP. The relative order of PPs is typically MAN-NER, AT-LOCATION and AT-TIME (Hawkins, 1999) , and this feature captures this idea without requiring temporal or local annotation.", |
| "cite_spans": [ |
| { |
| "start": 272, |
| "end": 287, |
| "text": "(Hawkins, 1999)", |
| "ref_id": "BIBREF11" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "PPs having quotes are more likely to encode a MANNER, the chunk of text between quotes being the manner.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "For example, use in \"very modest amounts\" (wsj 0003, 10) and reward with \"page bonuses\" (wsj 0012, 8).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "head-np indicates the head noun of the NP that attaches to the preposition to form the PP. It is retrieved by selecting the last noun in the NP. Certain nouns are more likely to indicate a MAN-NER than others. This feature captures the domain restriction. For nouns, only non temporal abstract objects and states can encode a MAN-NER. Some examples of positive instances are haul in the guests' [honor], lift in two [stages], win at any [cost] , plunge against the [mark] and ease with little [fanfare]. However, counterexamples can be found as well: say through his [spokesman] and do over the [counter] .", |
| "cite_spans": [ |
| { |
| "start": 437, |
| "end": 443, |
| "text": "[cost]", |
| "ref_id": null |
| }, |
| { |
| "start": 595, |
| "end": 604, |
| "text": "[counter]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Verb-pas indicates if a verb is in passive voice. In that case, a PP starting with by is much more likely to encode an AGENT than a MAN-NER. For example, compare (1) \"When the fruit is ripe, it [falls] y from the tree [by itself] PP ,\" he says. (wsj 0300, 23); and (2) Four of the planes [were purchased] y [by International Lease] PP from Singapore Airlines in a [. . . ] transaction (wsj 0243, 3) . In the first example a MANNER holds; in the second an AGENT.", |
| "cite_spans": [ |
| { |
| "start": 364, |
| "end": 372, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 373, |
| "end": 398, |
| "text": "transaction (wsj 0243, 3)", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "Given the sentence Kalipharma is a New Jerseybased pharmaceuticals concern that [sells products [under the Purepac label] PP ] VP . (wsj 0023, 1), the features are: {parent-node:VP, next-node:-, numpp-bef:0, num-pp-aft:0, first-word:under, first-POStag:IN, first-prep:under, POS-tag-bef:NNS, POStag-aft:DT, word-aft:the, has-rb:no, has-quotes:no, head-np-lemma:label, head-is-last:yes, head-hascap:yes, verb:sells, verb-lemma:sell, verb-pas:no}, and it is a positive instance.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Adverbial Phrases", |
| "sec_num": null |
| }, |
| { |
| "text": "As a learning algorithm we use a Naive Bayes classifier, well known for its simplicity and yet good performance. We trained our models with the training corpus using 10-fold cross validation, and used the held-out portion to tune the feature set and adjust parameters. More features than the ones depicted were tried, but we only report the final set. For example, named entity recognition and flags indicating the presence of AT-LOCATION and AT-TIME relations for the verb were tried, but they did not bring any significant improvement. Table 6 summarizes the results obtained. We report results only on the test corpus, which corresponds to instances not seen before and therefore they are a honest estimation of the performance. The improvement brought by subsets of features and statistical significance tests are also reported. We test the significance of the difference in performance between two feature sets i and j on a set of ins instances with the Z-score test, where", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 538, |
| "end": 545, |
| "text": "Table 6", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "z = abs(err i ,err j ) \u03c3 d", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": ", err k is the error made using set k, and", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "\u03c3 d = err i (1\u2212err i ) ins + err j (1\u2212err j ) ins .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "ADVPs The full set of features yields a F-measure of 0.815. The three specialized features (3, 4 and 5) are responsible for an improvement of .168 in the F-measure. This difference in performance yields a Z-score of 7.1, which indicates that it is statistically significant.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "PPs All the features proposed yield a F-measure of 0.693. The novel features specialized in MANNER detection from PPs (in bold letters in Table 5 ) bring an improvement of 0.059, which again is significant. The Z-score is 2.35, i.e., the difference in performance is statistically significant with a confidence greater than 97.5%. Thus, adding the specialized features is justified.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 138, |
| "end": 145, |
| "text": "Table 5", |
| "ref_id": "TABREF8" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Experimental Results", |
| "sec_num": "7.1" |
| }, |
| { |
| "text": "The mapping of words to semantic classes is data-driven and decisions were taken so that the overall accuracy is high. However, mistakes are made. Given We want to [see] y the market from the inside, the underlined PP encodes a MANNER and the mapping proposed (Table 1) does not map inside to ntao. Similarly, given Like their cohorts in political consulting, the litigation advisers [encourage] y their clients [. . . ] , the underlined text encodes a MANNER and yet cohorts is subsumed by social group.n.1 and therefore is not mapped to ntao.", |
| "cite_spans": [ |
| { |
| "start": 412, |
| "end": 420, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 260, |
| "end": 269, |
| "text": "(Table 1)", |
| "ref_id": "TABREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Error Analysis", |
| "sec_num": "7.2" |
| }, |
| { |
| "text": "The model proposed for MANNER detection makes mistakes as well. For ADVPs, if the main adverb has not been seen during training, chances of detecting MANNER are low. For example, the classifier fails to detect the following MANNER relations: [. . . ] which together own about [. . . ] (wsj 0671, 1); and who has ardently supported [. . . ] (wsj 1017, 26) even though ardently is present in the dictionary and ends in -ly;", |
| "cite_spans": [ |
| { |
| "start": 242, |
| "end": 250, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Error Analysis", |
| "sec_num": "7.2" |
| }, |
| { |
| "text": "For PPs, some errors are due to the Prop-Bank annotation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Error Analysis", |
| "sec_num": "7.2" |
| }, |
| { |
| "text": "For example, in Shearson Lehman Hutton began its coverage of the company with favorable ratings. (wsj 2061, 57), the underlined text is annotated as ARG2, even though it does encode a MANNER. Our model correctly detects a MANNER but it is counted as a mistake. Manners encoded by under and at are rarely detected, as in that have been consolidated in federal court under U.S. District Judge Milton Pollack (wsj 1022.mrg, 10).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Error Analysis", |
| "sec_num": "7.2" |
| }, |
| { |
| "text": "To the best of our knowledge, there have not been much efforts to detect MANNER alone. Girju et al. (2003) , present a supervised approach for ADVP similar to the one reported in this paper, yielding a F-measure of .665. Our augmented feature set obtains a F-measure of .815, clearly outperforming their method (Z-test, confidence > 97.5%). Moreover, ADVPs only represent 45.3% of MANNER as a semantic role in PropBank. We also have presented a model to detect MANNER encoded by a PP, the other big chunk of MANNER (44.6%) in PropBank.", |
| "cite_spans": [ |
| { |
| "start": 87, |
| "end": 106, |
| "text": "Girju et al. (2003)", |
| "ref_id": "BIBREF7" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Comparison with Previous Work", |
| "sec_num": "8" |
| }, |
| { |
| "text": "Complete systems for Semantic Role Labeling perform poorly when detecting MANNER; the Top-10 systems in CoNLL-2005 shared task 3 obtained F-measures ranging from .527 to .592. We have trained our models using the training data provided by the task organizers (using the Charniak parser syntactic information), and tested with the provided test set (test.wsj). Our models yield a Precision of .759 and Recall of .626 (F-measure .686), bringing a significant improvement over those systems (Z-test, confidence > 97.5%). When calculating recall, we take into account all MANNER in the test set, not only ADVPs and PPs whose fathers are S or VP (i.e. not only the ones our models are able to detect). This evaluation is done with exactly the same data provided from the task organizers for both training and test.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Comparison with Previous Work", |
| "sec_num": "8" |
| }, |
| { |
| "text": "Unlike typical semantic role labelers, our features do not include rich syntactic information (e.g. syntactic path from verb to the argument). Instead, we only require the value of the parent and in the case of PPs, the sibling node. When repeating the CoNLL-2005 Shared Task training and test using gold syntactic information, the F-measure obtained is .714, very close to the .686 obtained with Charniak syntactic trees (not significant, confidence > 97.5%).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Comparison with Previous Work", |
| "sec_num": "8" |
| }, |
| { |
| "text": "Even though syntactic parsers achieve a good performance, they make mistakes and the less our models rely on them, the better.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Comparison with Previous Work", |
| "sec_num": "8" |
| }, |
| { |
| "text": "MANNER can combine with other semantic relations in order to reveal implicit relations that otherwise would be missed. The basic idea is to compose MANNER with other relations in order to infer another MANNER. A necessary condition for combining MANNER with another relation R is the compatibility of RANGE(MNR) with DOMAIN(R) or RANGE(R) with DOMAIN(MNR). The extended definition (Section 3) allows to quickly determine if two relations are compatible (Blanco et al., 2010) .", |
| "cite_spans": [ |
| { |
| "start": 453, |
| "end": 474, |
| "text": "(Blanco et al., 2010)", |
| "ref_id": "BIBREF1" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "The new MANNER is automatically inferred by humans when reading, but computers need an explicit representation. Consider the following example: [. . . ] (wsj 0118, 48) . PropBank states the basic annotation between brackets: via computers is the MANNER and to buy the basket [. . . ] the PURPOSE of the place orders event. We propose to combine these two relations in order to come up with the new relation MNR(via computers, buy the basket [. . . ] ). This relation is obvious when reading the sentence, so it is omitted by the writer. However, any semantic representation of text needs as much semantics as possible explicitly stated.", |
| "cite_spans": [ |
| { |
| "start": 144, |
| "end": 152, |
| "text": "[. . . ]", |
| "ref_id": null |
| }, |
| { |
| "start": 275, |
| "end": 283, |
| "text": "[. . . ]", |
| "ref_id": null |
| }, |
| { |
| "start": 441, |
| "end": 449, |
| "text": "[. . . ]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 153, |
| "end": 167, |
| "text": "(wsj 0118, 48)", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "This claim is supported by several PropBank examples: (1) The classics have [zoomed] y [in price] MNR [to meet the competition] PRP , and . . . (wsj 0071, 9) and (2) . . . the government [curtailed] y production [with land-idling programs] MNR [to reduce price-depressing surpluses] PRP (wsj 0113, 12).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "In both examples, PropBank encodes the MANNER and PURPOSE for event y indicated with brackets. After combining both relations, two new MANNER arise: MNR(in price, meet the competition) and MNR(with land-idling programs, reduce price-depressing surpluses).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "Out of 237 verbs having in PropBank both PUR-POSE and MANNER annotation, the above inference method yields 189 new valid MANNER not present in PropBank (Accuracy .797).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "MANNER and other relations. MANNER does not combine with relations such as CAUSE, AT-LOCATION or AT-TIME. For example, given And they continue [anonymously] ", |
| "cite_spans": [ |
| { |
| "start": 143, |
| "end": 156, |
| "text": "[anonymously]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Composing MANNER with PURPOSE", |
| "sec_num": "9" |
| }, |
| { |
| "text": "We have presented a supervised method for the automatic discovery of MANNER. Our approach is simple and outperforms previous work. Our models specialize in detecting the most common pattern encoding MANNER. By doing so we are able to specialize our feature sets and outperform previous approaches that followed the idea of using dozens of features, most of them potentially useless, and letting a complicated machine learning algorithm decide the actual useful features.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "10" |
| }, |
| { |
| "text": "We believe that each relation or role has its own unique characteristics and capturing them improves performance. We have shown this fact for MANNER by examining examples, considering the kind of arguments that can be part of the domain and range, and considering theoretical works (Hawkins, 1999) .", |
| "cite_spans": [ |
| { |
| "start": 282, |
| "end": 297, |
| "text": "(Hawkins, 1999)", |
| "ref_id": "BIBREF11" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "10" |
| }, |
| { |
| "text": "We have shown performance using both gold syntactic trees and the output from the Charniak parser, and there is not a big performance drop. This is mainly due to the fact that we do not use deep syntactic information in our feature sets.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "10" |
| }, |
| { |
| "text": "The combination of MANNER and PURPOSE opens up a novel paradigm to perform semantic inference. We envision a layer of semantics using a small set of basic semantic relations and inference mechanisms on top of them to obtain more semantics on demand. Combining semantic relations in order to obtain more relation is only one of the possible inference methods.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "10" |
| }, |
| { |
| "text": "Penn TreeBank, file wsj 0027, sentence 10.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "http://verbs.colorado.edu/\u02dcmpalmer/projects/ace/PBguide lines.pdf, page 26.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "http://www.lsi.upc.es/\u02dcsrlconll/st05/st05.html", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "The Berkeley FrameNet Project", |
| "authors": [ |
| { |
| "first": "Collin", |
| "middle": [ |
| "F" |
| ], |
| "last": "Baker", |
| "suffix": "" |
| }, |
| { |
| "first": "Charles", |
| "middle": [ |
| "J" |
| ], |
| "last": "Fillmore", |
| "suffix": "" |
| }, |
| { |
| "first": "John", |
| "middle": [ |
| "B" |
| ], |
| "last": "Lowe", |
| "suffix": "" |
| } |
| ], |
| "year": 1998, |
| "venue": "Proceedings of the 17th international conference on Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Collin F. Baker, Charles J. Fillmore, and John B. Lowe. 1998. The Berkeley FrameNet Project. In Proceed- ings of the 17th international conference on Computa- tional Linguistics, Montreal, Canada.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Composition of Semantic Relations: Model and Applications", |
| "authors": [ |
| { |
| "first": "Eduardo", |
| "middle": [], |
| "last": "Blanco", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Hakki", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Cankaya", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Moldovan", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceedings of the 23rd International Conference on Computational Linguistics (COLING 2010)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Eduardo Blanco, Hakki C. Cankaya, and Dan Moldovan. 2010. Composition of Semantic Relations: Model and Applications. In Proceedings of the 23rd International Conference on Computational Linguistics (COLING 2010), Beijing, China.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Introduction to the CoNLL-2005 shared task: semantic role labeling", |
| "authors": [ |
| { |
| "first": "Xavier", |
| "middle": [], |
| "last": "Carreras", |
| "suffix": "" |
| }, |
| { |
| "first": "Llu\u00eds", |
| "middle": [], |
| "last": "M\u00e0rquez", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "CONLL '05: Proceedings of the Ninth Conference on Computational Natural Language Learning", |
| "volume": "", |
| "issue": "", |
| "pages": "152--164", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Xavier Carreras and Llu\u00eds M\u00e0rquez. 2005. Introduction to the CoNLL-2005 shared task: semantic role label- ing. In CONLL '05: Proceedings of the Ninth Confer- ence on Computational Natural Language Learning, pages 152-164, Morristown, NJ, USA.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "Incremental cue phrase learning and bootstrapping method for causality extraction using cue phrase and word pair probabilities", |
| "authors": [ |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Du", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Chang", |
| "suffix": "" |
| }, |
| { |
| "first": "S", |
| "middle": [], |
| "last": "Key", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Choi", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "Information Processing & Management", |
| "volume": "42", |
| "issue": "3", |
| "pages": "662--678", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Du S. Chang and Key S. Choi. 2006. Incremen- tal cue phrase learning and bootstrapping method for causality extraction using cue phrase and word pair probabilities. Information Processing & Management, 42(3):662-678.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Unsupervised Discovery of Generic Relationships Using Pattern Clusters and its Evaluation by Automatically Generated SAT Analogy Questions", |
| "authors": [ |
| { |
| "first": "Dmitry", |
| "middle": [], |
| "last": "Davidov", |
| "suffix": "" |
| }, |
| { |
| "first": "Ari", |
| "middle": [], |
| "last": "Rappoport", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of ACL-08: HLT", |
| "volume": "", |
| "issue": "", |
| "pages": "692--700", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Dmitry Davidov and Ari Rappoport. 2008. Unsuper- vised Discovery of Generic Relationships Using Pat- tern Clusters and its Evaluation by Automatically Gen- erated SAT Analogy Questions. In Proceedings of ACL-08: HLT, pages 692-700, Columbus, Ohio.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "Fully Unsupervised Discovery of Concept-Specific Relationships by Web Mining", |
| "authors": [ |
| { |
| "first": "Dmitry", |
| "middle": [], |
| "last": "Davidov", |
| "suffix": "" |
| }, |
| { |
| "first": "Ari", |
| "middle": [], |
| "last": "Rappoport", |
| "suffix": "" |
| }, |
| { |
| "first": "Moshe", |
| "middle": [], |
| "last": "Koppel", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "232--239", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Dmitry Davidov, Ari Rappoport, and Moshe Koppel. 2007. Fully Unsupervised Discovery of Concept- Specific Relationships by Web Mining. In Proceed- ings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 232-239, Prague, Czech Republic.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Automatic Labeling Of Semantic Roles", |
| "authors": [ |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "Gildea", |
| "suffix": "" |
| }, |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "Jurafsky", |
| "suffix": "" |
| } |
| ], |
| "year": 2002, |
| "venue": "Computational Linguistics", |
| "volume": "28", |
| "issue": "", |
| "pages": "245--288", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Daniel Gildea and Daniel Jurafsky. 2002. Automatic La- beling Of Semantic Roles. Computational Linguistics, 28:245-288.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Discovery of Manner Relations and Their Applicability to Question Answering", |
| "authors": [ |
| { |
| "first": "Roxana", |
| "middle": [], |
| "last": "Girju", |
| "suffix": "" |
| }, |
| { |
| "first": "Manju", |
| "middle": [], |
| "last": "Putcha", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Moldovan", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Proceedings of the ACL 2003 Workshop on Multilingual Summarization and Question Answering", |
| "volume": "", |
| "issue": "", |
| "pages": "54--60", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Roxana Girju, Manju Putcha, and Dan Moldovan. 2003. Discovery of Manner Relations and Their Applicabil- ity to Question Answering. In Proceedings of the ACL 2003 Workshop on Multilingual Summarization and Question Answering, pages 54-60, Sapporo, Japan.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Automatic Discovery of Part-Whole Relations", |
| "authors": [ |
| { |
| "first": "Roxana", |
| "middle": [], |
| "last": "Girju", |
| "suffix": "" |
| }, |
| { |
| "first": "Adriana", |
| "middle": [], |
| "last": "Badulescu", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Moldovan", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "Computational Linguistics", |
| "volume": "32", |
| "issue": "1", |
| "pages": "83--135", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Roxana Girju, Adriana Badulescu, and Dan Moldovan. 2006. Automatic Discovery of Part-Whole Relations. Computational Linguistics, 32(1):83-135.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "SemEval-2007 Task 04: Classification of Semantic Relations between Nominals", |
| "authors": [ |
| { |
| "first": "Roxana", |
| "middle": [], |
| "last": "Girju", |
| "suffix": "" |
| }, |
| { |
| "first": "Preslav", |
| "middle": [], |
| "last": "Nakov", |
| "suffix": "" |
| }, |
| { |
| "first": "Vivi", |
| "middle": [], |
| "last": "Nastase", |
| "suffix": "" |
| }, |
| { |
| "first": "Stan", |
| "middle": [], |
| "last": "Szpakowicz", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [], |
| "last": "Turney", |
| "suffix": "" |
| }, |
| { |
| "first": "Deniz", |
| "middle": [], |
| "last": "Yuret", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the Fourth International Workshop on Semantic Evaluations", |
| "volume": "", |
| "issue": "", |
| "pages": "13--18", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Roxana Girju, Preslav Nakov, Vivi Nastase, Stan Sz- pakowicz, Peter Turney, and Deniz Yuret. 2007. SemEval-2007 Task 04: Classification of Semantic Relations between Nominals. In Proceedings of the Fourth International Workshop on Semantic Evalua- tions (SemEval-2007), pages 13-18, Prague, Czech Republic.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Semantic role labeling via FrameNet, VerbNet and Prop-Bank", |
| "authors": [ |
| { |
| "first": "Ana", |
| "middle": [ |
| "M" |
| ], |
| "last": "Giuglea", |
| "suffix": "" |
| }, |
| { |
| "first": "Alessandro", |
| "middle": [], |
| "last": "Moschitti", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "929--936", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ana M. Giuglea and Alessandro Moschitti. 2006. Se- mantic role labeling via FrameNet, VerbNet and Prop- Bank. In Proceedings of the 21st International Con- ference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, pages 929-936, Morristown, NJ, USA.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "The relative order of prepositional phrases in English: Going beyond Manner-Place-Time", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "John", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Hawkins", |
| "suffix": "" |
| } |
| ], |
| "year": 1999, |
| "venue": "Language Variation and Change", |
| "volume": "11", |
| "issue": "03", |
| "pages": "231--266", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "John A. Hawkins. 1999. The relative order of prepo- sitional phrases in English: Going beyond Manner- Place-Time. Language Variation and Change, 11(03):231-266.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Automatic Acquisition of Hyponyms from Large Text Corpora", |
| "authors": [ |
| { |
| "first": "Marti", |
| "middle": [ |
| "A" |
| ], |
| "last": "Hearst", |
| "suffix": "" |
| } |
| ], |
| "year": 1992, |
| "venue": "Proceedings of the 14th International Conference on Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "539--545", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Marti A. Hearst. 1992. Automatic Acquisition of Hy- ponyms from Large Text Corpora. In Proceedings of the 14th International Conference on Computational Linguistics, pages 539-545.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Knowledge Representation and the Semantics of Natural Language", |
| "authors": [ |
| { |
| "first": "Hermann", |
| "middle": [], |
| "last": "Helbig", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hermann Helbig. 2005. Knowledge Representation and the Semantics of Natural Language. Springer.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Detecting Semantic Relations between Named Entities in Text Using Contextual Features", |
| "authors": [ |
| { |
| "first": "Toru", |
| "middle": [], |
| "last": "Hirano", |
| "suffix": "" |
| }, |
| { |
| "first": "Yoshihiro", |
| "middle": [], |
| "last": "Matsuo", |
| "suffix": "" |
| }, |
| { |
| "first": "Genichiro", |
| "middle": [], |
| "last": "Kikui", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics, Demo and Poster Sessions", |
| "volume": "", |
| "issue": "", |
| "pages": "157--160", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Toru Hirano, Yoshihiro Matsuo, and Genichiro Kikui. 2007. Detecting Semantic Relations between Named Entities in Text Using Contextual Features. In Pro- ceedings of the 45th Annual Meeting of the Associa- tion for Computational Linguistics, Demo and Poster Sessions, pages 157-160, Prague, Czech Republic.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Semantic Role Labeling: An Introduction to the Special Issue", |
| "authors": [ |
| { |
| "first": "Llu\u00eds", |
| "middle": [], |
| "last": "M\u00e0rquez", |
| "suffix": "" |
| }, |
| { |
| "first": "Xavier", |
| "middle": [], |
| "last": "Carreras", |
| "suffix": "" |
| }, |
| { |
| "first": "Kenneth", |
| "middle": [ |
| "C" |
| ], |
| "last": "Litkowski", |
| "suffix": "" |
| }, |
| { |
| "first": "Suzanne", |
| "middle": [], |
| "last": "Stevenson", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Computational Linguistics", |
| "volume": "34", |
| "issue": "2", |
| "pages": "145--159", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Llu\u00eds M\u00e0rquez, Xavier Carreras, Kenneth C. Litkowski, and Suzanne Stevenson. 2008. Semantic Role Label- ing: An Introduction to the Special Issue. Computa- tional Linguistics, 34(2):145-159.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "Lymba's PowerAnswer 4 in TREC", |
| "authors": [ |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Moldovan", |
| "suffix": "" |
| }, |
| { |
| "first": "Christine", |
| "middle": [], |
| "last": "Clark", |
| "suffix": "" |
| }, |
| { |
| "first": "Mitchell", |
| "middle": [], |
| "last": "Bowden", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the Sixteenth Text REtrieval Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Dan Moldovan, Christine Clark, and Mitchell Bowden. 2007. Lymba's PowerAnswer 4 in TREC 2007. In Proceedings of the Sixteenth Text REtrieval Confer- ence (TREC 2007).", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Semantic Classification of Noun Phrases Using Web Counts and Learning Algorithms", |
| "authors": [ |
| { |
| "first": "Paul", |
| "middle": [], |
| "last": "Nulty", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Proceedings of the ACL 2007 Student Research Workshop", |
| "volume": "", |
| "issue": "", |
| "pages": "79--84", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Paul Nulty. 2007. Semantic Classification of Noun Phrases Using Web Counts and Learning Algorithms. In Proceedings of the ACL 2007 Student Research Workshop, pages 79-84, Prague, Czech Republic.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "The Proposition Bank: An Annotated Corpus of Semantic Roles", |
| "authors": [ |
| { |
| "first": "Martha", |
| "middle": [], |
| "last": "Palmer", |
| "suffix": "" |
| }, |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "Gildea", |
| "suffix": "" |
| }, |
| { |
| "first": "Paul", |
| "middle": [], |
| "last": "Kingsbury", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Computational Linguistics", |
| "volume": "31", |
| "issue": "1", |
| "pages": "71--106", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Martha Palmer, Daniel Gildea, and Paul Kingsbury. 2005. The Proposition Bank: An Annotated Cor- pus of Semantic Roles. Computational Linguistics, 31(1):71-106.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "Extraction of Entailed Semantic Relations Through Syntax-Based Comma Resolution", |
| "authors": [ |
| { |
| "first": "Vivek", |
| "middle": [], |
| "last": "Srikumar", |
| "suffix": "" |
| }, |
| { |
| "first": "Roi", |
| "middle": [], |
| "last": "Reichart", |
| "suffix": "" |
| }, |
| { |
| "first": "Mark", |
| "middle": [], |
| "last": "Sammons", |
| "suffix": "" |
| }, |
| { |
| "first": "Ari", |
| "middle": [], |
| "last": "Rappoport", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Roth", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of ACL-08: HLT", |
| "volume": "", |
| "issue": "", |
| "pages": "1030--1038", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Vivek Srikumar, Roi Reichart, Mark Sammons, Ari Rap- poport, and Dan Roth. 2008. Extraction of Entailed Semantic Relations Through Syntax-Based Comma Resolution. In Proceedings of ACL-08: HLT, pages 1030-1038, Columbus, Ohio.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Interactive semantic analysis of Clause-Level Relationships", |
| "authors": [ |
| { |
| "first": "Ken", |
| "middle": [], |
| "last": "Barker Szpakowicz", |
| "suffix": "" |
| }, |
| { |
| "first": "Stan", |
| "middle": [], |
| "last": "Barker", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Szpakowicz", |
| "suffix": "" |
| } |
| ], |
| "year": 1995, |
| "venue": "Proceedings of the Second Conference of the Pacific Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "22--30", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Barker Szpakowicz, Ken Barker, and Stan Szpakowicz. 1995. Interactive semantic analysis of Clause-Level Relationships. In Proceedings of the Second Confer- ence of the Pacific Association for Computational Lin- guistics, pages 22-30.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "Automatic Discovery of Intentions in Text and its Application to Question Answering", |
| "authors": [ |
| { |
| "first": "Marta", |
| "middle": [], |
| "last": "Tatu", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Proceedings of the ACL Student Research Workshop", |
| "volume": "", |
| "issue": "", |
| "pages": "31--36", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Marta Tatu. 2005. Automatic Discovery of Intentions in Text and its Application to Question Answering. In Proceedings of the ACL Student Research Workshop, pages 31-36, Ann Arbor, Michigan.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "Expressing Implicit Semantic Relations without Supervision", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Peter", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Turney", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics", |
| "volume": "", |
| "issue": "", |
| "pages": "313--320", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Peter D. Turney. 2006. Expressing Implicit Semantic Relations without Supervision. In Proceedings of the 21st International Conference on Computational Lin- guistics and 44th Annual Meeting of the Association for Computational Linguistics, pages 313-320, Syd- ney, Australia.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "num": null, |
| "text": "MANNER is close in meaning with several other relations, specifically INSTRUMENT, AT-LOCATION and AT-TIME.", |
| "type_str": "figure", |
| "uris": null |
| }, |
| "TABREF0": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "-STRUMENT of the broke event. Other relations that may be confused as MANNER include AT-LOCATION and AT-TIME, like in [The dog jumped] x [over the fence] y and [John used to go] x [regularly] y ." |
| }, |
| "TABREF1": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "Class Rule situation state || event state POStag=verb || isHypo(state.n.4) event (POStag=verb && in(verb events)) || (POStag=noun && !animate object && (isHypo(phenomenon.n.1) || isHypo(event.n.1) || in(noun events))) animate object livingNE || (POStag=noun && ((isHypo(entity.n.1) && !isHypo(thing.n.9) && !isHypo(anticipation.n.4)) || isHypo(social group.n.1))) livingNE neType=(human | organization | country | town | province | other-loc) quality POStag=(adverb | gerund) || headPP=(with | without) non temporal abstract object abstract object && !temporal abstract object neType=money || isHypo(thing.n.9) || (!isHypo(social group.n.1) && (isHypo(abstraction.n.6 | psychological feature.n.1 | possession.n.2 | event.n.1 | state.n.4 | group.n.1 | act.n.2))) temporal TemporalNE || isHypo(time period.n.1) || isHypo(time.n.5) temporalNE ne-type=(date | time)" |
| }, |
| "TABREF2": { |
| "type_str": "table", |
| "content": "<table><tr><td>Synt. pattern</td><td>#Occ. %Occ.</td><td>File, #sent</td><td>Example Sentence</td></tr><tr><td colspan=\"3\">ADVP 45.3% wsj 0039, 24 Other 3559 240 3.1% wsj 1337, 0</td><td/></tr></table>", |
| "html": null, |
| "num": null, |
| "text": "Mapping for the semantic classes used for defining DOMAIN(MNR) and RANGE(MNR). This story line might [resonate] y [more strongly] ADVP if Mr. Lane had as strong a presence in front of the camera as he does behind it. PP 3499 44.6% wsj 2451, 0 NBC may yet find a way to [take] y a passive, minority interest in a program-maker [without violating the rules] PP . RB 286 3.6% wsj 0052, 3 Backe is [a [[closely] RB [held] y ] ADJP media firm] NP run by former CBS Inc. President Jon Backe. S 148 1.9% wsj 1217, 25 Salomon [posted] y an unexpectedly big gain in quarterly earnings, [aided by its securities trading and investments banking activities] S . NP 120 1.5% wsj 0100, 21 [. . . ] he [graduated] y [Phi Beta Kappa] NP from the University of Kentucky at age 18, after spending only 2 1/2 years in college. Tokyo stocks [closed] y [firmer] ADJP Monday, with the Nikkei index making its fifth consecutive daily gain." |
| }, |
| "TABREF3": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "Syntactic patterns encoding MANNER in PropBank, number of occurrences, and examples. A total of 7,852 MANNER relations are encoded in PropBank between a single node in the syntactic tree and a verb. In all examples, MNR(x, y) holds, where x is the text underlined. Syntactic annotation comes straight from the Penn TreeBank. a verb; MANNER relations expressed by trace chains identifying coreference and split arguments are ignored. This way, we consider 7,852 MANNER out of the total of the 8,037 PropBank annotates. Because ADVPs and PPs represent 90% of MANNER relations, in this paper we focus exclusively on these two phrases." |
| }, |
| "TABREF4": { |
| "type_str": "table", |
| "content": "<table><tr><td colspan=\"2\">Parent #Occ.</td><td>Phrase File, #sent</td><td>Example</td><td>Sentence</td></tr><tr><td>S</td><td colspan=\"2\">215 ADVP wsj 0044, 6 [[339 PP wsj 2454, 9 [[</td><td/></tr></table>", |
| "html": null, |
| "num": null, |
| "text": "VP 3306 ADVP wsj 2341, 23 The company [was [officially] ADVP [merged] y with Bristol-Myers Co. earlier this month] VP . 3107 PP wsj 2320, 7 This is something P&G [would [do] y [with or without Kao] PP ] VP , says Mr. Zurkuhlen. Virtually word by word] ADVP , the notes [matched] y questions and answers on the social-studies section of the test the student was taking.] S Under the laws of the land] PP , the ANC [remains] y an illegal organization, and its headquarters are still in Lusaka, Zambia.] S ADJP 17 ADVP wsj 1057, 85 [. . . ] ABC touted \"Call to Glory,\" but the military drama was [[missing] y [in action] PP ] ADJP within weeks. 4 PP wsj 2431, 14 Two former ministers [were] y [[so heavily] ADVP implicated] ADJP in the Koskotas affair that PASOK members of Parliament voted [. . . ] PP 9 ADVP wsj 1249, 24 In Japan, by contrast, companies tend to develop their talent and [promote] y [from [within] PP ] PP . 9 PP wsj 1505, 30 London share prices were [influenced]" |
| }, |
| "TABREF5": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "" |
| }, |
| "TABREF6": { |
| "type_str": "table", |
| "content": "<table><tr><td colspan=\"2\">No. Feature</td><td>Values</td><td>Explanation</td></tr><tr><td>1</td><td>parent-node</td><td>{S, VP}</td><td>syntactic node of ADVP's parent</td></tr><tr><td>2</td><td>num-leaves</td><td>N</td><td>number of words in ADVP</td></tr><tr><td>3</td><td>adverb</td><td colspan=\"2\">{often, strongly, . . . } main adverb of ADVP</td></tr><tr><td>4</td><td>dictionary</td><td>{yes, no}</td><td>is adverb is in dictionary?</td></tr><tr><td>5</td><td colspan=\"2\">ends-with-ly {yes, no}</td><td>does adverb end in -ly?</td></tr><tr><td>6</td><td>POS-tag-bef</td><td>POStags</td><td>POS tag word before adverb</td></tr><tr><td>7</td><td>POS-tag-aft</td><td>POStags</td><td>POS tag word after adverb</td></tr><tr><td>8</td><td>verb</td><td>{assigned, go, . . . }</td><td>main verb the ADVP attaches to</td></tr><tr><td>9</td><td>distance</td><td>N</td><td>number of words between adverb and verb</td></tr></table>", |
| "html": null, |
| "num": null, |
| "text": "and the participating systems in" |
| }, |
| "TABREF7": { |
| "type_str": "table", |
| "content": "<table><tr><td colspan=\"2\">No. Feature</td><td>Values</td><td>Explanation</td></tr><tr><td>1</td><td>parent-node</td><td>{S, VP}</td><td>syntactic node of PP's parent</td></tr><tr><td>2</td><td>next-node</td><td>{NP, SBAR, , . . . }</td><td>syntactic node of sibling to the right of PP</td></tr><tr><td>3</td><td>num-pp-bef</td><td>N</td><td>number of sibling before PP which are PP</td></tr><tr><td>4</td><td>num-pp-aft</td><td>N</td><td>number of sibling after PP which are PP</td></tr><tr><td>5</td><td>first-word</td><td>{with, after, . . . }</td><td>first word in PP</td></tr><tr><td>6</td><td colspan=\"2\">first-POS-tag POStags</td><td>first POS tag in PP</td></tr><tr><td>7</td><td>first-prep</td><td>{by, on, . . . }</td><td>first preposition in PP</td></tr><tr><td>8</td><td>POS-tag-bef</td><td>POStags</td><td>POS tag before first-word</td></tr><tr><td>9</td><td>POS-tag-aft</td><td>POStags</td><td>POs tag after first-word</td></tr><tr><td colspan=\"2\">10 word-aft</td><td>{one, their, . . . }</td><td>word after first-word</td></tr><tr><td colspan=\"2\">11 has-rb</td><td>{yes, no}</td><td>does the PP contain an adverb?</td></tr><tr><td colspan=\"2\">12 has-quotes</td><td>{yes, no}</td><td>does the PP have any quotes?</td></tr><tr><td colspan=\"2\">13 head-14 head-is-last</td><td>{yes, no}</td><td>is head-np the last word of the sentence?</td></tr><tr><td colspan=\"2\">15 head-has-cap</td><td>{yes, no}</td><td>does the PP have a capitalized word?</td></tr><tr><td colspan=\"2\">16 verb</td><td colspan=\"2\">{approved, fly, . . . } verb the PP attaches to</td></tr><tr><td colspan=\"2\">17 verb-lemma</td><td>{approve, be, . . . }</td><td>verb lemma the PP attaches to</td></tr><tr><td colspan=\"2\">18 verb-pas</td><td>{yes, no}</td><td>is verb in passive voice?</td></tr></table>", |
| "html": null, |
| "num": null, |
| "text": "Features used for extracting MANNER from ADVPs, their values and explanation. Features 4 and 5 are specialized on MANNER detection. , year, . . . } head of the NP whose parent is the PP" |
| }, |
| "TABREF8": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "Features used for extracting MANNER from PPs, their values and explanation. Features in bold letters are new and specialized on detecting MANNER from PPs." |
| }, |
| "TABREF9": { |
| "type_str": "table", |
| "content": "<table><tr><td colspan=\"3\">Phrase #MNR Feat. Set</td><td colspan=\"2\">#MNR retrieved #MNR correct</td><td>P</td><td>R</td><td>F</td></tr><tr><td>ADVP</td><td>678</td><td>1,2,6-9 all</td><td>908 757</td><td>513 585</td><td colspan=\"2\">.565 .757 .647 .773 .863 .815</td></tr><tr><td>PP</td><td>705</td><td>1,2,5,6,8-10,16,17 all</td><td>690 713</td><td>442 491</td><td colspan=\"2\">.641 .627 .634 .689 .696 .693</td></tr></table>", |
| "html": null, |
| "num": null, |
| "text": "Table 6: Results obtained during testing for different sets of features." |
| }, |
| "TABREF10": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "the traders [place] y orders [via computers] MNR [to buy the basket of stocks . . . ] PRP" |
| }, |
| "TABREF11": { |
| "type_str": "table", |
| "content": "<table/>", |
| "html": null, |
| "num": null, |
| "text": "x,MNR [attacking] y CIA Director William Webster [for being too accommodating to the committee] z,CAU (wsj 0590, 27), there is no relation between x and z. Similarly, given [In the tower] x,LOC , five men and women [pull] y [rhythmically] z,MNR on ropes attached to [. . . ] (wsj 0089, 5) and [In May] x,TMP , the two companies, [through their jointly owned holding company] z,MNR , Temple, [offered] y [. . . ] (wsj 0063, 3), no connection exists between x and z." |
| } |
| } |
| } |
| } |
