| { |
| "paper_id": "Y14-1029", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T13:44:25.870633Z" |
| }, |
| "title": "Annotating article errors in Spanish learner texts: design and evaluation of an annotation scheme", |
| "authors": [ |
| { |
| "first": "Mar\u00eda", |
| "middle": [], |
| "last": "Del", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Prefectural University", |
| "location": { |
| "addrLine": "Nagakute-shi Aichi", |
| "postCode": "1522-3, 480-1198", |
| "settlement": "Ibaragasama", |
| "country": "Japan" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Pilar", |
| "middle": [], |
| "last": "Valverde Iba\u00f1ez", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Prefectural University", |
| "location": { |
| "addrLine": "Nagakute-shi Aichi", |
| "postCode": "1522-3, 480-1198", |
| "settlement": "Ibaragasama", |
| "country": "Japan" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Akira", |
| "middle": [], |
| "last": "Ohtani", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "Gakuin University", |
| "location": { |
| "addrLine": "2-36-1 Kishibe-minami, Suita-shi Osaka", |
| "postCode": "564-8511", |
| "country": "Japan" |
| } |
| }, |
| "email": "ohtani@ogu.ac.jp" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "Annotating a corpus with error information is a challenging task. This paper describes the design, evaluation and refinement of an annotation scheme for Spanish article errors in learner data, so that future work on corpus annotation and automatic article error detection can progress. To evaluate reliability, 300 noun phrases with definite, indefinite and zero article have been tagged by four annotators. We analysed different types of disagreement, presented suggestions to increase reliability and applied the refined annotation scheme to create a gold-standard annotation.", |
| "pdf_parse": { |
| "paper_id": "Y14-1029", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "Annotating a corpus with error information is a challenging task. This paper describes the design, evaluation and refinement of an annotation scheme for Spanish article errors in learner data, so that future work on corpus annotation and automatic article error detection can progress. To evaluate reliability, 300 noun phrases with definite, indefinite and zero article have been tagged by four annotators. We analysed different types of disagreement, presented suggestions to increase reliability and applied the refined annotation scheme to create a gold-standard annotation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "The annotation of learner texts with error information is necessary for linguistic research as well as for the development of language learning applications using natural language processing (NLP) techniques. While much efforts have concentrated on English, it is necessary to develop learner corpora and tools for other foreign languages like Spanish. This is the most commonly studied foreign language in the United States and the second most studied foreign language -after English-in many other countries. Overall, it is estimated that nearly 20 million people are studying Spanish as a foreign language (Instituto Cervantes, 2013) . However, learner corpora and tools for this language are scarce (Lozano and Mendikoetxea, 2013; Nazar and Renau, 2012; del Pilar Valverde and Ohtani, 2012; Wanner et al., 2013) . The goal of this paper is to define an annotation scheme that is suitable for reliable Spanish ar-ticle error annotation, so that future work on corpus annotation and automatic article error detection can progress.", |
| "cite_spans": [ |
| { |
| "start": 608, |
| "end": 635, |
| "text": "(Instituto Cervantes, 2013)", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 702, |
| "end": 733, |
| "text": "(Lozano and Mendikoetxea, 2013;", |
| "ref_id": "BIBREF20" |
| }, |
| { |
| "start": 734, |
| "end": 756, |
| "text": "Nazar and Renau, 2012;", |
| "ref_id": "BIBREF23" |
| }, |
| { |
| "start": 757, |
| "end": 793, |
| "text": "del Pilar Valverde and Ohtani, 2012;", |
| "ref_id": null |
| }, |
| { |
| "start": 794, |
| "end": 814, |
| "text": "Wanner et al., 2013)", |
| "ref_id": "BIBREF29" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Automatic detection of errors has focused on function words such as articles (Izumi et al., 2004; Han et al., 2006; Felice and Pulman, 2008b; Gamon et al., 2008; Yi et al., 2008) , prepositions (Felice and Pulman, 2008a) and particles (Dickinson, 2008; Oyama and Matsumoto, 2010) . Function words are the most frequent words in any language, and they are also a very common source of mistakes for learners.", |
| "cite_spans": [ |
| { |
| "start": 77, |
| "end": 97, |
| "text": "(Izumi et al., 2004;", |
| "ref_id": "BIBREF16" |
| }, |
| { |
| "start": 98, |
| "end": 115, |
| "text": "Han et al., 2006;", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 116, |
| "end": 141, |
| "text": "Felice and Pulman, 2008b;", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 142, |
| "end": 161, |
| "text": "Gamon et al., 2008;", |
| "ref_id": "BIBREF12" |
| }, |
| { |
| "start": 162, |
| "end": 178, |
| "text": "Yi et al., 2008)", |
| "ref_id": "BIBREF30" |
| }, |
| { |
| "start": 194, |
| "end": 220, |
| "text": "(Felice and Pulman, 2008a)", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 235, |
| "end": 252, |
| "text": "(Dickinson, 2008;", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 253, |
| "end": 279, |
| "text": "Oyama and Matsumoto, 2010)", |
| "ref_id": "BIBREF24" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "As for error annotation, one of the main difficulties is reliability. For some learner errors, like number and gender agreement, rules are clearly defined. Other kind of errors, like article or preposition presence and choice, are harder to annotate because native speakers differ widely with respect to what is acceptable usage. For article and noun number selection, for example, in Lee et al. (2009) raters found more than one valid construction for more than 18% of noun phrases.", |
| "cite_spans": [ |
| { |
| "start": 385, |
| "end": 402, |
| "text": "Lee et al. (2009)", |
| "ref_id": "BIBREF18" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "To address this problem, we experiment with a preliminary annotation scheme for article errors, analyse the form disagreement among annotators takes, and refine the annotation scheme according to it. The paper is organized as follows. In section 2 we briefly summarize the linguistic properties of Spanish articles. In section 3 we explain an experiment carried out with a preliminary annotation scheme on article error annotation. In section 4 we examine the sources of disagreement among the annotators and in 5 we summarize the recommendations for reliable annotation. Section 6 presents the conclusions. In Spanish, articles can be definite (as in English the) or indefinite (in English a/an), and their form changes according to the gender and number of the noun they complement, as shown in Table 1 Article usage is complex because it is the result of the interaction of pragmatic, semantic, syntactic and lexical factors. Taxonomies of article use are abundant in the literature, targeted towards learners (Butt and Benjamin, 2014) or linguists (Bosque and Demonte, 1999; RAE, 2009) . Basically, the main function of articles is to indicate the relationship between the nominal expressions and the entities to which the speakers refer by means of such expressions (Bosque and Demonte, 1999) . For example, among other usages, we use the definite to generalize, that is, to refer to a whole class of things or people, as in (1) and to refer to something that is identifiable to the listener, as in (2). In (2), Maria's son must be identifiable for the listener either because a) Maria has only one son, or b) we have talked about him before. We use the indefinite to refer to any object of a particular class, as in (3), and we use no article when we are talking about an indefinite amount of something, as in (4) (examples from ).", |
| "cite_spans": [ |
| { |
| "start": 1013, |
| "end": 1038, |
| "text": "(Butt and Benjamin, 2014)", |
| "ref_id": "BIBREF3" |
| }, |
| { |
| "start": 1052, |
| "end": 1078, |
| "text": "(Bosque and Demonte, 1999;", |
| "ref_id": null |
| }, |
| { |
| "start": 1079, |
| "end": 1089, |
| "text": "RAE, 2009)", |
| "ref_id": null |
| }, |
| { |
| "start": 1271, |
| "end": 1297, |
| "text": "(Bosque and Demonte, 1999)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 797, |
| "end": 804, |
| "text": "Table 1", |
| "ref_id": "TABREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "(1)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Los hijos dan muchos disgustos. 'Children cause a great deal of trouble.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "(2) El hijo de Mar\u00eda tiene dos a\u00f1os. 'Maria's son is two years old.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "(3) Tener un hijo es lo mejor que te puede pasar en esta vida. 'Having a child is the best thing that can happen in life.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "(4) No tengo hijos pero tengo sobrinos. 'I do not have children but I have nephews.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "With regard to syntactic factors, for example two or more coordinated nouns should have their own article if they refer to different things: un gato y un perro, \"a cat and dog\" (un gato y perro suggests a cross between a cat and a dog) (Butt and Benjamin, 2014) . As for semantic factors, there are many rules which require specific knowledge. For example, place names usually have no article (M\u00e9xico). For some of them the article is optional ((el) Per\u00fa) or depends on the context (el M\u00e9xico de los mexicanos, \"Mexicans' Mexico\"), while the definite is obligatory for rivers, mountains, seas and oceans (el Mediterr\u00e1neo). Other rules exist for numbers, proper nouns, names of languages, days of the week, etc.", |
| "cite_spans": [ |
| { |
| "start": 236, |
| "end": 261, |
| "text": "(Butt and Benjamin, 2014)", |
| "ref_id": "BIBREF3" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Finally, there exist many set phrases and idioms which require definite (e.g. con el objetivo de 'with the objective of'), indefinite (por una parte, 'on the one hand') or zero article (e.g. a corto plazo, 'in the short run').", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Definite articles are the most frequent word in Spanish. In Davies (2005) frequency list the definite article is the most frequent type and the indefinite article is the 7th most frequent. In 9 billion words Spanish TenTen corpus (Jakub\u00ed\u010dek et al., 2013) the definite is also the most frequent type and the indefinite is the 6th. Approximately one out of every ten words in this corpus are articles.", |
| "cite_spans": [ |
| { |
| "start": 60, |
| "end": 73, |
| "text": "Davies (2005)", |
| "ref_id": "BIBREF6" |
| }, |
| { |
| "start": 230, |
| "end": 254, |
| "text": "(Jakub\u00ed\u010dek et al., 2013)", |
| "ref_id": "BIBREF17" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Difficulties for learners", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Articles are also one of the most frequent grammatical errors, specially for speakers of languages that do not have articles like Chinese, Japanese, Korean or Russian. For speakers of Japanese, Fern\u00e1ndez (1997) found 2.2 article errors per 100 words in a 4433 words sample. 2 In addition to that, this type of error diminishes as proficiency increases, but it tends to fossilize. The difficulty of the article system of Spanish may be comparable to English. McEnery et al. (2006) found that articles were the most difficult to acquire for Japanese learn-ers of English, since even proficient learners had not achieved the acquisition rate of 90%. Therefore, we decided to use Japanese learners' texts to develop our annotation scheme.", |
| "cite_spans": [ |
| { |
| "start": 194, |
| "end": 210, |
| "text": "Fern\u00e1ndez (1997)", |
| "ref_id": "BIBREF11" |
| }, |
| { |
| "start": 274, |
| "end": 275, |
| "text": "2", |
| "ref_id": null |
| }, |
| { |
| "start": 458, |
| "end": 479, |
| "text": "McEnery et al. (2006)", |
| "ref_id": "BIBREF22" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Difficulties for learners", |
| "sec_num": "2.2" |
| }, |
| { |
| "text": "Annotation of learner errors is a challenging task for several reasons. First, learner sentences often contain interacting surrounding errors which can make the understanding of the meaning of the sentence quite difficult. Second, for some errors like number and gender agreement there are clear-cut rules about what is grammatical. But for other kind of errors, like article or preposition presence and choice, rules are usually not clearly defined, so in some cases more than one article choice may be acceptable. And third, in some cases more textual context or world knowledge may be needed to be able to determine the correct article usage.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "As a result, inter-annotator agreement for error annotations can be relatively low. This issue has been put forward by the NLP community, that has found difficulties for evaluating error detection systems (Chodorow et al., 2012) , but it has not received much attention in the learner corpus linguistic field. Several measures can be taken to address the varying number of corrections and levels of acceptability a sentence can have.", |
| "cite_spans": [ |
| { |
| "start": 205, |
| "end": 228, |
| "text": "(Chodorow et al., 2012)", |
| "ref_id": "BIBREF4" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "With regard to the number of possible analysis a sentence can receive, most error-annotated learner corpora permit only one tag per error. However, the \"single correct construction\" approach has been questioned and in recent annotation efforts there is a tendency to allow the inclusion of several alternative codes for the same item (L\u00fcdeling et al., 2005; Boyd, 2010; Lee et al., 2012; Rozovskaya and Roth, 2010) . However, it is unattainable to list all possible interpretations for every error, so this is done only \"when there is doubt\".", |
| "cite_spans": [ |
| { |
| "start": 334, |
| "end": 357, |
| "text": "(L\u00fcdeling et al., 2005;", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 358, |
| "end": 369, |
| "text": "Boyd, 2010;", |
| "ref_id": "BIBREF2" |
| }, |
| { |
| "start": 370, |
| "end": 387, |
| "text": "Lee et al., 2012;", |
| "ref_id": "BIBREF19" |
| }, |
| { |
| "start": 388, |
| "end": 414, |
| "text": "Rozovskaya and Roth, 2010)", |
| "ref_id": "BIBREF27" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "With regard to the level of confidence in the annotators' judgments, some projects include global measures of inter-annotator agreement (Rozovskaya and Roth, 2010; Lee et al., 2012) but annotated corpora do not explicitly provide confidence levels for every error. Only in some annotation experiments the annotators are asked to indicate their level of confidence for every item (as \"low\" or \"high\") (Tetreault and Chodorow, 2008) .", |
| "cite_spans": [ |
| { |
| "start": 136, |
| "end": 163, |
| "text": "(Rozovskaya and Roth, 2010;", |
| "ref_id": "BIBREF27" |
| }, |
| { |
| "start": 164, |
| "end": 181, |
| "text": "Lee et al., 2012)", |
| "ref_id": "BIBREF19" |
| }, |
| { |
| "start": 400, |
| "end": 430, |
| "text": "(Tetreault and Chodorow, 2008)", |
| "ref_id": "BIBREF28" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "We carry out an experiment on annotation of article errors with the following objectives:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "1. Calculate inter-annotator agreement.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "2. Analyse the types and sources of disagreement, to find out which are the main difficulties the annotators face when annotating article errors in learner texts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "3. Based on this experience, refine the guidelines and annotation scheme for error annotation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experiment", |
| "sec_num": "3" |
| }, |
| { |
| "text": "We used learners' texts written by 4th grade Japanese students of Spanish with an intermediate level of proficiency, at Aichi Prefectural University. A teacher of Spanish as a Foreign Language extracted sentences containing at least one article error from these texts, 50 sentences for each kind of article (definite, indefinite and zero article). The same number of sentences, but with at least one correct article usage, was then collected from the same texts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Data collection", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "In every sentence only one highlighted noun phrase had to be annotated. The distribution of the resulting 300 sentences is as Table 2 shows. Correct 50 50 50 150 Incorrect 50 50 50 150 Total 100 100 100 300 Table 2 : Number of noun phrases and articles they contain", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 126, |
| "end": 133, |
| "text": "Table 2", |
| "ref_id": null |
| }, |
| { |
| "start": 141, |
| "end": 230, |
| "text": "Correct 50 50 50 150 Incorrect 50 50 50 150 Total 100 100 100 300 Table 2", |
| "ref_id": "TABREF2" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Data collection", |
| "sec_num": "3.1" |
| }, |
| { |
| "text": "The 300 noun phrases were tagged by four annotators. The annotators were two experts (teachers of Spanish as a Foreign Language, who correct learners' texts on a regular basis), which we will call E1 and E2, and two non-experts (native speakers of Spanish with higher education but without experience in corpus annotation), which we will call NE1 and NE2. They all annotated the same noun phrase in the same sentences, but presented in different orders, using a Microsoft Excel spreadsheet. Annotators were provided with the target sentence plus the preceding and the following sentence, which they could resort to if they needed more context. If the target sentence was at the beginning or end of paragraph or text in the original text, no context was provided (a \"beginning or end of paragraph or text\" mark was inserted instead).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Preliminary annotation scheme", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "They were asked to classify the noun phrase into one of the categories shown in Table 3 . We are only concerned with article presence and choice, so we did not tag malformation (e.g. spelling or agreement) and order errors.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 80, |
| "end": 87, |
| "text": "Table 3", |
| "ref_id": "TABREF3" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Preliminary annotation scheme", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Missing (definite) AD Missing (indefinite) AI Extraneous E Confusion C Article is correct OK Difficult to judge NC Difficult to judge (NC) It was expected that the annotators would some times be unsure about the acceptability of article usage in a given sentence, or unable to determine the most likely correction. We opted for allowing only one tag per sentence, but not forcing the annotators to mark the article usage as \"right\" or \"wrong\" and instead gave the possibility of marking sentences as \"difficult to judge\", as Han et al. (2006) . We we wanted the annotators to correct the sentences only when they were sure about their decision, and not forcing them to make a best guess, which could lower inter-annotator agreement. Later we could look at the sentences marked as problematic, as 14, and analyse what they have in common.", |
| "cite_spans": [ |
| { |
| "start": 525, |
| "end": 542, |
| "text": "Han et al. (2006)", |
| "ref_id": "BIBREF13" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Preliminary annotation scheme", |
| "sec_num": "3.2" |
| }, |
| { |
| "text": "Tables 4 and 5 show the confusion matrices for expert and non-expert annotations. Observed agreement 4 is 0.79 for expert annotators and 0.76 for nonexperts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Inter-annotator agreement", |
| "sec_num": "4" |
| }, |
| { |
| "text": "However, using observed agreement to measure reliability does not take into account agreement that is due to chance and hence is not a good measure of reliability. Therefore, an analysis using Cohen's Kappa statistic (Cohen, 1960) chance agreement would equate to 0. For the whole set of sentences (300, correct or incorrect), interannotator agreement for experts was found to be Kappa = 0.71 (p < 0.001), 95% CI (0.65, 0.77), and for non-experts it was 0.68 (p < 0.001), 95% CI (0.62, 0.75), which indicates substantial agreement. If we exclude the 45 sentences marked as\"difficult to judge\" by at least one annotator, kappa is 0.85 and 0.73 respectively. If we exclude 97 sentences tagged as \"correct\" by the four of them, remaining only sentences where at least one annotator considers there is an error, kappa is 0.62 and 0.58. If we exclude both sentences marked as NC by at least one annotator and sentences marked as OK by four annotators (remaining only 159 sentences) kappa is 0.79 and 0.61. In the following sections we examine different types of disagreement: disagreement due to the annotators (4.1), due to the annotation scheme (4.2) and genuine disagreement (4.3), and propose some measures to reduce it.", |
| "cite_spans": [ |
| { |
| "start": 217, |
| "end": 230, |
| "text": "(Cohen, 1960)", |
| "ref_id": "BIBREF5" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Inter-annotator agreement", |
| "sec_num": "4" |
| }, |
| { |
| "text": "The difference between experts and non-experts' reliability is due to the fact that non-experts make more slips than experts, and they are also less conservative when they correct texts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "In the data we find at least five mistakes (there can be more which we cannot detect), all by nonexpert annotators: in four sentences they tag for a missing article a noun phrase which already contains one article, as (9) (C|C|AD|OK||OK), and in another one they tag for an extraneous article error a noun phrase without article. 9En Guatemala, la gente que tiene alta ense\u00f1anza piensa que \"voseo\" es una norma culta. 'In Guatemala, people who have higher education think that \"voseo\" is an educated norm.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "To prevent this kind of mistakes, any annotation project should automatically constrain the tags the annotators can use depending on the input (e.g. if there is already an article preceding a noun phrase, do not allow the \"missing\" error tag). Table 6 shows the error tags a noun phrase can receive depending on the article it contains.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 244, |
| "end": 251, |
| "text": "Table 6", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Error tag Definite Indefinite 0 article AD x AI", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "x C", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "x x E", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "x x Table 6 : Error tags a noun phrase can receive depending on the type of article it contains", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 4, |
| "end": 11, |
| "text": "Table 6", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "In addition to that, even though non-experts are supposed to be less confident about the acceptability of sentences because pointing out errors in a text is a task for which they have no previous experience, in fact they are less cautious when they correct texts. For example, in (10) (OK|OK|E|E||OK)) experts consider the article is acceptable, while nonexperts classify it as an extraneous article.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotators expertise: experts vs non-experts", |
| "sec_num": "4.1" |
| }, |
| { |
| "text": "Segundo,ahora ya no es imprescindible usar la coca para los objetivos antiguos,como para alivia de dolor o anestesia [...] . 'Second, now it is no longer necessary to use the coca for the ancient purposes, like pain relieve or anaesthetic [...] .'", |
| "cite_spans": [ |
| { |
| "start": 117, |
| "end": 122, |
| "text": "[...]", |
| "ref_id": null |
| }, |
| { |
| "start": 239, |
| "end": 244, |
| "text": "[...]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(10)", |
| "sec_num": null |
| }, |
| { |
| "text": "This bias explains why, for example, NE1 uses the tag \"difficult to judge\" only one time (0.3%), while E2 uses it almost once every 10 sentences (9.3%), and non-experts use the tag \"extraneous article\" (specially for definite articles) more frequently than experts (23.5% vs. 12.2% of times).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(10)", |
| "sec_num": null |
| }, |
| { |
| "text": "Principle of minimal change Part of the variability on annotators' rigour could be reduced by giving clear guidelines about the optimum level of intervention in the texts. In this regard, we advocate for following a principle of minimal change: so we should not mark as errors the sentences where the learner choice is acceptable, even if the learner choice is not the best choice, that is, the goal of the annotator should be to produce an acceptable rather than a perfect result (e.g. Hana et al. (2010) ), When the input is incomprehensible and the annotator cannot make a decision, it should be left without annotation.", |
| "cite_spans": [ |
| { |
| "start": 487, |
| "end": 505, |
| "text": "Hana et al. (2010)", |
| "ref_id": "BIBREF14" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(10)", |
| "sec_num": null |
| }, |
| { |
| "text": "In relation to that, annotators should be informed about the halo effect, by which the judgement of a sentence as acceptable or unacceptable is influenced by our overall impression of previous sentences. In other words, one is more likely to find errors in a text if this text already contains other errors. Experts (teachers of a foreign language) are trained on evaluation methods and they are aware of the importance of reliability in students' evaluation. They know how external factors (e.g. the halo effect and contrast effect) can have a negative impact and what can be done to reduce it. However, non-experts lack this training and are not aware of the challenges faced to perform a fair evaluation -annotation.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(10)", |
| "sec_num": null |
| }, |
| { |
| "text": "We find some disagreements are due to the design of the preliminary annotation scheme, specially concerning the tags \"difficult to judge\" (NC) and \"confusion error\" (C).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "The tag \"difficult to judge\" With regard to the reliability of the 6 tags used for annotation (Table 3) , \"difficult to judge\" is the one that causes more disagreement: most of the times (67.7%) it is used by only one of the four annotators, and it is never used by three or four annotators in the same sentence. On the contrary, the rest of tags have a much higher agreement: on average, they are used by the four annotators 63.2% of the times, by three 19.9%, by two 9.2% and by one 7.7% of times.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 94, |
| "end": 103, |
| "text": "(Table 3)", |
| "ref_id": "TABREF3" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Therefore, this tag should at most be used to filter out problematic sentences, which annotators cannot comprehend, and not for proper annotation of sentences.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "We advocate for not using this tag and instead set clear principles in the annotation guidelines specifying what the annotators should do when they are not confident about the error analysis of a sentence.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "The tag \"confusion error\" We found there was ambiguity in the guidelines about the meaning of this tag: in principle, it refers to the confusion between definite and indefinite articles but annotators also use it to indicate the confusion between an article and another type of determiner.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "Indeed, learners frequently confuse the indefinite article with the indefinite determiner alguno 'some', when they refer to an indefinite amount of things, as in (11) (C|C|OK|OK||CD). 11Los hispanos est\u00e1n aumentando r\u00e1pidamente y la poblaci\u00f3n est\u00e1 centrada en unos estados. 'Hispanics are increasing rapidly and the population is concentrated in some states.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "To include this kind of error in the annotation, we should break down the tag into two: confusion between definite and indefinite article (CA) and confusion between article and another type of determiner (CD).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Disagreement due to the annotation scheme", |
| "sec_num": "4.2" |
| }, |
| { |
| "text": "As explained in section 2, article presence and choice can be determined by several factors. In our data, it mainly depends on pragmatic factors (69.0% of noun phrases), followed by lexicosemantic (20.7%) and syntactic factors (10.3%).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Genuine disagreement", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Leaving aside sentences tagged as acceptable by four annotators, agreement is higher when the article choice depends on lexico-semantic factors (k = 0.835 for experts and 0.780 for non-experts) and lower with pragmatic factors ((k = 0.514 for experts and 0.496 for non-experts). Syntactic factors seem to be in between (k = 0.750 for experts and 0.523 for non-experts), although their low frequency makes the figures less reliable. Therefore, more care should be paid to pragmatic distinctions.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Genuine disagreement", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Specifically, disagreement is more likely in noun phrases where two pragmatic interpretations (and article choices) are possible, and annotators choose one of the alternatives in an inconsistent manner ( \u00a7 4.3.1 and 4.3.2) . Disagreement can also be due to a lack of the world knowledge that is needed to be able to determine the correct article usage ( \u00a7 4.3.3) . As for syntactic and lexico-semantic factors ( \u00a7 4.3.4), disagreement occurs because annotators do not have a good knowledge about the existing prescriptive rules about article usage.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 202, |
| "end": 222, |
| "text": "( \u00a7 4.3.1 and 4.3.2)", |
| "ref_id": null |
| }, |
| { |
| "start": 352, |
| "end": 362, |
| "text": "( \u00a7 4.3.3)", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Genuine disagreement", |
| "sec_num": "4.3" |
| }, |
| { |
| "text": "Frequently both definite and zero article are acceptable for the same noun phrase. This happens when the noun phrase can refer to a whole class of things or people in general (definite article) or to an indefinite amount of something (zero article), as explained in 2. This distinction frequently does not change the meaning of the sentence significantly and in fact some languages with articles like English usually use the zero article in both cases.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "When both pragmatic interpretations are possible for a given sentence, annotators unevenly choose one of them: some annotators tag the noun phrase for a missing article in (12) (OK|AD|AD|OK||OK) while they tag it for extraneous article in (13) (E |NC|OK|E||OK), even though in both sentences both the definite article and the zero article are acceptable, so the learner's choice should be left unchanged. 12Los pol\u00edticos hablan en p\u00fablico y manifiestan sus opiniones con el objeto de conseguir votos de ciudadanos [...] 'Politicians talk in public and show their opinion with a view to get votes from the citizens [...] .'", |
| "cite_spans": [ |
| { |
| "start": 514, |
| "end": 519, |
| "text": "[...]", |
| "ref_id": null |
| }, |
| { |
| "start": 614, |
| "end": 619, |
| "text": "[...]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "Concretamente los cursos que consiguieron participantes japoneses y que ofrecen los certificados oficiales como IMEC(Instituto de Medicina China) continuar\u00e1n existiendo [...] . 'Specifically the courses which obtained Japanese participants and offer official certificates like IMEC (Chinese Medicine Institute) will continue existing [...] .'", |
| "cite_spans": [ |
| { |
| "start": 169, |
| "end": 174, |
| "text": "[...]", |
| "ref_id": null |
| }, |
| { |
| "start": 334, |
| "end": 339, |
| "text": "[...]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "This distinction is specially problematic with plural nouns: in noun phrases with a plural nominal head, agreement by four annotators is less frequent (43.2%) than with singular nouns (66.7%)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "2 (2, N = 299) = 18.9, p < 0.001. Therefore, more care should be paid in the annotation of plural nouns.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "If the noun is singular and uncountable, we find the same ambiguous pragmatic distinction as with plural nouns, as in (14) (NC|NC|AD|E||OK), which is tagged as \"difficult to judge\" by some annotators and \"extraneous\" by others (the AD tag is a lapsus). 14El problema es demanda de la coca\u00edna. 'The problem is demand of cocaine.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "In conclusion, according to the principle of minimal change, when both the definite and the zero article are acceptable, we should leave the learners' choice unchanged.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Definite article or zero article", |
| "sec_num": "4.3.1" |
| }, |
| { |
| "text": "Some times annotators agree in considering a noun phrase as unacceptable but they do not agree in the type of correction. This can happen when the learner wrongly uses a definite article, as in (15) (E|C|C|E||E/CA), and the annotators propose different corrections for it because the noun phrase can refer to an indefinite amount of something (zero article) or any object of a particular class (indefinite). 15En cambio, la coca\u00edna tiene el efecto t\u00f3xico. 'On the contrary, cocaine has a toxic effect.' Only in these cases, we allow adding two error tags (E/CA or E/CD) to the noun phrase.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Indefinite article or zero article", |
| "sec_num": "4.3.2" |
| }, |
| { |
| "text": "In some sentences, annotators have insufficient extra-linguistic knowledge to be able to determine the right article usage. For example, in (16) (OK|E|E|E||OK) the annotator needs to know whether in Nagoya there are only nine interesting and touristy places (definite article) or there are more than nine (no article).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "World knowledge", |
| "sec_num": "4.3.3" |
| }, |
| { |
| "text": "Sale cada treinta minutos aproximadamente desde la estaci\u00f3n de Nagoya y paran en los nueve sitios muy interesantes y tur\u00edsticos, por ejemplo El castillo de Nagoya. 'It runs approximately every thirty minutes from Nagoya station and stops in nine very interesting and touristy places, for example Nagoya Castle.'", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(16)", |
| "sec_num": null |
| }, |
| { |
| "text": "If the learner's choice is acceptable in some context, as in (16), we do not mark it as wrong. If the learner's choice is not acceptable, we tag the noun phrase as usual.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "(16)", |
| "sec_num": null |
| }, |
| { |
| "text": "Unlike article usage governed by pragmatic factors, which is subject to interpretation by the annotator, for article usage determined by syntactic and lexico-semantic constraints there exist some linguistic norms about what is considered correct and incorrect.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Syntactic and lexico-semantic rules", |
| "sec_num": "4.3.4" |
| }, |
| { |
| "text": "However, native speakers -even experts-do not have a deep knowledge about these rules and some times do not follow them. For example, in (17) (AD|AD|OK|OK||OK) experts marked as error an article usage that is actually accepted, while nonexperts tagged it right. It is the use of zero article between the preposition a ('to') and the relative pronoun que ('which') (RAE, 2006) . 17[...] el cap\u00edtulo 2 dice sobre el proceso del portu\u00f1ol y los problemas a que el portu\u00f1ol se enfrenta actualmente. ' [...] chapter 2 is about the portu\u00f1ol process and the problems that the portu\u00f1ol confronts nowadays.' Therefore, to determine the acceptability of article usage, annotators should not rely only on their intuition as native speakers but also consult existing rules and recommendations published in reference dictionaries and grammars as RAE (2006) and RAE (2009) .", |
| "cite_spans": [ |
| { |
| "start": 364, |
| "end": 375, |
| "text": "(RAE, 2006)", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 496, |
| "end": 501, |
| "text": "[...]", |
| "ref_id": null |
| }, |
| { |
| "start": 832, |
| "end": 842, |
| "text": "RAE (2006)", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 847, |
| "end": 857, |
| "text": "RAE (2009)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Syntactic and lexico-semantic rules", |
| "sec_num": "4.3.4" |
| }, |
| { |
| "text": "After examining the sources of disagreement in the annotation experiment, we added the following principles to the annotation scheme:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "1. It is not recommended to use a tag like NC, \"difficult to judge\", because it has the lowest reliability. Therefore, we recommend simply not annotating the noun phrase if it impossible to determine the acceptability of the article usage. We did not find any case like that in our data from students with an intermediate level of Spanish.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "2. Tags should inform us about the type of error and about the correction. This was true for the \"add definite\", \"add indefinite\" and \"delete\" tags, since we indicate which article we should add (definite or indefinite), and we know which article is deleted. The preliminary \"confusion\" error tag should be broken down into two tags to indicate confusion between definite and indefinite article (CA), and confusion between article and another type of determiner (CD).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "3. Follow the principle of minimal change: the sentences should be acceptable rather than perfect. When more than one article choice including the learner's one is acceptable, we leave the learner's choice as correct. The pair definite article-zero article is the most interchangeable (in many sentences both are correct), so annotators should pay attention not to change the learner choice when it is correct.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "4. When the learner choice is not acceptable and there are two equally good corrections, we allow double annotation. We found this mainly happens when the learner wrongly uses a definite or indefinite, and the annotators doubt between an extraneous error (zero article) and a confusion error. Only in this cases, we allow double annotation with E and CA or CD tags. There is usually no ambiguity in the appropriate correction for a missing article: annotators usually agree whether a definite or indefinite is necessary (probably for this reason the zero article has a high inter-annotator agreement.)", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "5. Regarding article usage governed by syntactic and lexico-semantic factors, base annotation not only on annotators' intuitions but first on the rules about article usage published by respected institutions (RAE, 2006; RAE, 2009) .", |
| "cite_spans": [ |
| { |
| "start": 208, |
| "end": 219, |
| "text": "(RAE, 2006;", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 220, |
| "end": 230, |
| "text": "RAE, 2009)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "6. When more world knowledge is needed to judge a sentence as correct or incorrect, we do not correct it if the learner's choice is acceptable in some context.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Following these criteria, we have revised the error tags given by the annotators for every sentence and made a decision about the most acceptable tag. The articles in the resulting gold standard set are distributed as Table 7 shows.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 218, |
| "end": 225, |
| "text": "Table 7", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Suggestions for reliable annotation", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Definite Table 7 : Frequency of error tags in the gold standard per type of article (absolute frequency or %)", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 9, |
| "end": 16, |
| "text": "Table 7", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "Tag", |
| "sec_num": null |
| }, |
| { |
| "text": "Despite the small size of the corpus study, some tendencies are observed in the 300 noun phrases written by Japanese learners:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Tag", |
| "sec_num": null |
| }, |
| { |
| "text": "1. The most frequent error regarding the definite article is extraneous use (83.7%): learners overuse it frequently probably because it is the most frequent article (and word) in Spanish.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Tag", |
| "sec_num": null |
| }, |
| { |
| "text": "2. When zero article is used, the most likely error is omission of the definite article (86.9%), for the same reason.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Tag", |
| "sec_num": null |
| }, |
| { |
| "text": "3. When learners use an indefinite article, the errors they commit are more evenly distributed. Confusion with a definite article or another type of determiner happens in 54.8% of cases and extraneous use in 42.9%.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Tag", |
| "sec_num": null |
| }, |
| { |
| "text": "Although article errors have been annotated in a number of small-scale studies, to date there has not been any study about article error annotation and inter-annotator agreement in Spanish learner texts.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "6" |
| }, |
| { |
| "text": "In this paper we have tested the results of an annotation scheme for article errors in a sample of learner texts written by Japanese learners. We have calculated agreement among four annotators (two experts and two non-experts) and have found kappa values between 0.85 and 0.62 for expert annotators and from 0.73 to 0.58 for non-experts, depending on the collection of sentences considered. The analysis of the disagreement among annotators has served us to find which are the main difficulties for annotators and to refine the annotation scheme according to it. Following more articulated guidelines we have revised the data to create a gold-standard. The data used for the experiment is available to all interested researchers upon request. We hope the work presented here will facilitate future corpus annotation and development of automatic article error detection systems.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusions", |
| "sec_num": "6" |
| }, |
| { |
| "text": "Spanish also has a definite article with neuter gender (lo), but its usage is quite different from the rest, so it will not be considered in this paper.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The most frequent grammatical error in her sample concerns the verb (3.2 verb tense errors per 100 words), followed by prepositions (2.8 per 100 words) and articles.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "For every example from the learner data, in parenthesis we indicate the tags by the four annotators, in the following order: E1|E2|NE1|NE2||gold standard. For more details about the gold standard version, see section 5.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Defined as the number of items on which annotators agree divided by the total number of items", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "This research was partially supported by kakenhi (25770207 and 24500189), Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science.We thank the students from Aichi Prefectural University who gave their permission to use their texts for this research.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgments", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "Students' Basic Grammar of Spanish", |
| "authors": [ |
| { |
| "first": "Rosario", |
| "middle": [], |
| "last": "Alonso", |
| "suffix": "" |
| }, |
| { |
| "first": "Alejandro", |
| "middle": [], |
| "last": "Casta\u00f1eda", |
| "suffix": "" |
| }, |
| { |
| "first": "Pablo", |
| "middle": [], |
| "last": "Mart\u00ednez", |
| "suffix": "" |
| }, |
| { |
| "first": "Lourdes", |
| "middle": [], |
| "last": "Miguel", |
| "suffix": "" |
| }, |
| { |
| "first": "Jenaro", |
| "middle": [], |
| "last": "Ortega", |
| "suffix": "" |
| }, |
| { |
| "first": "Jos\u00e9", |
| "middle": [], |
| "last": "Ruiz", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rosario Alonso, Alejandro Casta\u00f1eda, Pablo Mart\u00ednez, Lourdes Miguel, Jenaro Ortega, and Jos\u00e9 Ruiz. 2013. Students' Basic Grammar of Spanish. Difusion.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Descriptive Grammar of Spanish Language. Espasa Calpe", |
| "authors": [], |
| "year": 1999, |
| "venue": "Spanish: Gram\u00e1tica descriptiva de la lengua espa\u00f1ola", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ignacio Bosque and Violeta Demonte, editors. 1999. Descriptive Grammar of Spanish Language. Espasa Calpe, (In Spanish: Gram\u00e1tica descriptiva de la lengua espa\u00f1ola).", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "EAGLE: an error-annotated corpus of beginning learner German", |
| "authors": [ |
| { |
| "first": "Adriane", |
| "middle": [], |
| "last": "Boyd", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceedings of LREC-10", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Adriane Boyd. 2010. EAGLE: an error-annotated cor- pus of beginning learner German. In Proceedings of LREC-10, Malta.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "A New Reference Grammar of Modern Spanish", |
| "authors": [ |
| { |
| "first": "John", |
| "middle": [], |
| "last": "Butt", |
| "suffix": "" |
| }, |
| { |
| "first": "Carmen", |
| "middle": [], |
| "last": "Benjamin", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "John Butt and Carmen Benjamin. 2014. A New Refer- ence Grammar of Modern Spanish. Routledge.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Problems in evaluating grammatical error detection systems", |
| "authors": [ |
| { |
| "first": "Martin", |
| "middle": [], |
| "last": "Chodorow", |
| "suffix": "" |
| }, |
| { |
| "first": "Markus", |
| "middle": [], |
| "last": "Dickinson", |
| "suffix": "" |
| }, |
| { |
| "first": "Ross", |
| "middle": [], |
| "last": "Israel", |
| "suffix": "" |
| }, |
| { |
| "first": "Joel", |
| "middle": [], |
| "last": "Tetreault", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of COLING 2012", |
| "volume": "", |
| "issue": "", |
| "pages": "611--628", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Martin Chodorow, Markus Dickinson, Ross Israel, and Joel Tetreault. 2012. Problems in evaluating gram- matical error detection systems. In Proceedings of COLING 2012, pages 611-628, Mumbai, Desember.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "A coefficient of agreement for nominal scales. Educational and Psychological Measurement", |
| "authors": [ |
| { |
| "first": "Jacob", |
| "middle": [], |
| "last": "Cohen", |
| "suffix": "" |
| } |
| ], |
| "year": 1960, |
| "venue": "", |
| "volume": "20", |
| "issue": "", |
| "pages": "37--46", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jacob Cohen. 1960. A coefficient of agreement for nom- inal scales. Educational and Psychological Measure- ment, 20(1):37-46.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "A Frequency Dictionary of Spanish: Core Vocabulary for Learners (CD)", |
| "authors": [ |
| { |
| "first": "Mark", |
| "middle": [], |
| "last": "Davies", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Mark Davies. 2005. A Frequency Dictionary of Spanish: Core Vocabulary for Learners (CD). Routledge.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Automatic detection of gender and number agreement errors in Spanish texts written by Japanese learners", |
| "authors": [], |
| "year": 2012, |
| "venue": "Proceedings of the 26th PACLIC", |
| "volume": "", |
| "issue": "", |
| "pages": "299--307", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Mar\u00eda del Pilar Valverde and Akira Ohtani. 2012. Au- tomatic detection of gender and number agreement er- rors in Spanish texts written by Japanese learners. In Proceedings of the 26th PACLIC, pages 299-307.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "Korean particle error detection via probabilistic parsing", |
| "authors": [ |
| { |
| "first": "Markus", |
| "middle": [], |
| "last": "Dickinson", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Automatic Analysis of Learner Language (AALL'08)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Markus Dickinson. 2008. Korean particle error detec- tion via probabilistic parsing. In Automatic Analysis of Learner Language (AALL'08).", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Automatic detection of preposition errors in learner writing", |
| "authors": [ |
| { |
| "first": "Rachele", |
| "middle": [], |
| "last": "De Felice", |
| "suffix": "" |
| }, |
| { |
| "first": "Stephen", |
| "middle": [ |
| "G" |
| ], |
| "last": "Pulman", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Automatic Analysis of Learner Language (AALL'08)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rachele De Felice and Stephen G. Pulman. 2008a. Automatic detection of preposition errors in learner writing. In Automatic Analysis of Learner Language (AALL'08).", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "A classifier-based approach to preposition and determiner error correction in L2 English", |
| "authors": [ |
| { |
| "first": "Rachele", |
| "middle": [], |
| "last": "De Felice", |
| "suffix": "" |
| }, |
| { |
| "first": "Stephen", |
| "middle": [ |
| "G" |
| ], |
| "last": "Pulman", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of the COLING 2008", |
| "volume": "", |
| "issue": "", |
| "pages": "169--176", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rachele De Felice and Stephen G. Pulman. 2008b. A classifier-based approach to preposition and deter- miner error correction in L2 English. In Proceedings of the COLING 2008, pages 169-176, Manchester, UK.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Interlanguage and Error Analysis in the Learning of Spanish as a Foreign Language. Edelsa", |
| "authors": [ |
| { |
| "first": "Soledad", |
| "middle": [], |
| "last": "Fern\u00e1ndez", |
| "suffix": "" |
| } |
| ], |
| "year": 1997, |
| "venue": "Spanish: Interlengua y an\u00e1lisis de errores en el aprendizaje del espa\u00f1ol como lengua extranjera)", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Soledad Fern\u00e1ndez. 1997. Interlanguage and Error Analysis in the Learning of Spanish as a Foreign Lan- guage. Edelsa, (In Spanish: Interlengua y an\u00e1lisis de errores en el aprendizaje del espa\u00f1ol como lengua ex- tranjera).", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Using contextual spell checker techniques and language modelling for ESL error correction", |
| "authors": [ |
| { |
| "first": "Michael", |
| "middle": [], |
| "last": "Gamon", |
| "suffix": "" |
| }, |
| { |
| "first": "Jianfeng", |
| "middle": [], |
| "last": "Gao", |
| "suffix": "" |
| }, |
| { |
| "first": "Chris", |
| "middle": [], |
| "last": "Brockett", |
| "suffix": "" |
| }, |
| { |
| "first": "Alex", |
| "middle": [], |
| "last": "Klementiev", |
| "suffix": "" |
| }, |
| { |
| "first": "William", |
| "middle": [ |
| "B" |
| ], |
| "last": "Dolan", |
| "suffix": "" |
| }, |
| { |
| "first": "Dimitry", |
| "middle": [], |
| "last": "Belenko", |
| "suffix": "" |
| }, |
| { |
| "first": "Lucy", |
| "middle": [], |
| "last": "Vanderwende", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of the International Joint Conference on Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "449--456", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Michael Gamon, Jianfeng Gao, Chris Brockett, Alex Klementiev, William B. Dolan, Dimitry Belenko, and Lucy Vanderwende. 2008. Using contextual spell checker techniques and language modelling for ESL error correction. In Proceedings of the International Joint Conference on Natural Language Processing, pages 449-456, Hyderabad, India.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "Detecting errors in English article usage by non-native speakers", |
| "authors": [ |
| { |
| "first": "Na-Rae", |
| "middle": [], |
| "last": "Han", |
| "suffix": "" |
| }, |
| { |
| "first": "Martin", |
| "middle": [], |
| "last": "Chodorow", |
| "suffix": "" |
| }, |
| { |
| "first": "Claudia", |
| "middle": [], |
| "last": "Leacock", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "Natural Language Engineering", |
| "volume": "12", |
| "issue": "2", |
| "pages": "115--129", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Na-Rae Han, Martin Chodorow, and Claudia Leacock. 2006. Detecting errors in English article usage by non-native speakers. Natural Language Engineering, 12(2):115-129.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Error-tagged learner corpus of Czech", |
| "authors": [ |
| { |
| "first": "Jirka", |
| "middle": [], |
| "last": "Hana", |
| "suffix": "" |
| }, |
| { |
| "first": "Alexandr", |
| "middle": [], |
| "last": "Rosen", |
| "suffix": "" |
| }, |
| { |
| "first": "Sva", |
| "middle": [], |
| "last": "", |
| "suffix": "" |
| }, |
| { |
| "first": "Barbora\u0161tindlov\u00e1", |
| "middle": [], |
| "last": "", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceedings of the Fourth Linguistic Annotation Workshop (ACL 2010)", |
| "volume": "", |
| "issue": "", |
| "pages": "11--19", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Jirka Hana, Alexandr Rosen, Sva, and Barbora\u0160tindlov\u00e1. 2010. Error-tagged learner corpus of Czech. In Pro- ceedings of the Fourth Linguistic Annotation Work- shop (ACL 2010), pages 11-19, Uppsala, Sweden, July.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "Spanish: a Living Language", |
| "authors": [ |
| { |
| "first": "Cervantes", |
| "middle": [], |
| "last": "Ic Instituto", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Spanish: El espa\u00f1ol: una lengua viva. Informe", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "IC Instituto Cervantes. 2013. Spanish: a Living Lan- guage. 2013 Report. Instituto Cervantes, (In Spanish: El espa\u00f1ol: una lengua viva. Informe 2013).", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "SST speech corpus of Japanese learners' English and automatic detection of learners' errors. International Computer Archive of", |
| "authors": [ |
| { |
| "first": "Emi", |
| "middle": [], |
| "last": "Izumi", |
| "suffix": "" |
| }, |
| { |
| "first": "Kiyotaka", |
| "middle": [], |
| "last": "Uchimoto", |
| "suffix": "" |
| }, |
| { |
| "first": "Hitoshi", |
| "middle": [], |
| "last": "Isahara", |
| "suffix": "" |
| } |
| ], |
| "year": 2004, |
| "venue": "Modern English Journal", |
| "volume": "28", |
| "issue": "", |
| "pages": "31--48", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Emi Izumi, Kiyotaka Uchimoto, and Hitoshi Isahara. 2004. SST speech corpus of Japanese learners' English and automatic detection of learners' errors. International Computer Archive of Modern English Journal, 28:31-48.", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Vojt\u011bch Kov\u00e1\u0159, Pavel Rychl\u00fd, and V\u00edt Suchomel", |
| "authors": [ |
| { |
| "first": "Milo\u0161", |
| "middle": [], |
| "last": "Jakub\u00ed\u010dek", |
| "suffix": "" |
| }, |
| { |
| "first": "Adam", |
| "middle": [], |
| "last": "Kilgarriff", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "7th International Corpus Linguistics Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Milo\u0161 Jakub\u00ed\u010dek, Adam Kilgarriff, Vojt\u011bch Kov\u00e1\u0159, Pavel Rychl\u00fd, and V\u00edt Suchomel. 2013. The TenTen corpus family. In 7th International Corpus Linguistics Con- ference.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "Human evaluation of article and noun number usage: Influences of context and construction variability", |
| "authors": [ |
| { |
| "first": "John", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| }, |
| { |
| "first": "Joel", |
| "middle": [], |
| "last": "Tetreault", |
| "suffix": "" |
| }, |
| { |
| "first": "Martin", |
| "middle": [], |
| "last": "Chodorow", |
| "suffix": "" |
| } |
| ], |
| "year": 2009, |
| "venue": "Proceedings of the Third Linguistic Annotation Workshop (LAW)", |
| "volume": "", |
| "issue": "", |
| "pages": "60--63", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "John Lee, Joel Tetreault, and Martin Chodorow. 2009. Human evaluation of article and noun number usage: Influences of context and construction variability. In Proceedings of the Third Linguistic Annotation Work- shop (LAW), pages 60-63, Suntec, Singapore.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "Developing learner corpus annotation for korean particle errors", |
| "authors": [ |
| { |
| "first": "Sun-Hee", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| }, |
| { |
| "first": "Markus", |
| "middle": [], |
| "last": "Dickinson", |
| "suffix": "" |
| }, |
| { |
| "first": "Ross", |
| "middle": [], |
| "last": "Israel", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of the Sixth Linguistic Annotation Workshop, LAW VI", |
| "volume": "", |
| "issue": "", |
| "pages": "129--133", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Sun-Hee Lee, Markus Dickinson, and Ross Israel. 2012. Developing learner corpus annotation for korean par- ticle errors. In Proceedings of the Sixth Linguistic An- notation Workshop, LAW VI, pages 129-133, Strouds- burg.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Learner corpora and second language acquisition: the design and collection of CEDEL2", |
| "authors": [ |
| { |
| "first": "Crist\u00f3bal", |
| "middle": [], |
| "last": "Lozano", |
| "suffix": "" |
| }, |
| { |
| "first": "Amaya", |
| "middle": [], |
| "last": "Mendikoetxea", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Automatic Treatment and Analysis of Learner Corpus Data. John Benjamins", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Crist\u00f3bal Lozano and Amaya Mendikoetxea. 2013. Learner corpora and second language acquisition: the design and collection of CEDEL2. In Ana D\u00edaz- Negrillo, Nicolas Ballier, and Paul Thompson, editors, Automatic Treatment and Analysis of Learner Corpus Data. John Benjamins, Amsterdam.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "Multi-level error annotation in learner corpora", |
| "authors": [ |
| { |
| "first": "Anke", |
| "middle": [], |
| "last": "L\u00fcdeling", |
| "suffix": "" |
| }, |
| { |
| "first": "Maik", |
| "middle": [], |
| "last": "Walter", |
| "suffix": "" |
| }, |
| { |
| "first": "Emil", |
| "middle": [], |
| "last": "Kroymann", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [], |
| "last": "Adolphs", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Proceedings of the Corpus Linguistics 2005 Conference", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anke L\u00fcdeling, Maik Walter, Emil Kroymann, and Pe- ter Adolphs. 2005. Multi-level error annotation in learner corpora. In Proceedings of the Corpus Linguis- tics 2005 Conference, Birmingham, United Kingdom, July.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "L2 acquisition of grammatical morphemes. In Corpusbased language studies. An advanced resource book", |
| "authors": [ |
| { |
| "first": "Tony", |
| "middle": [], |
| "last": "Mcenery", |
| "suffix": "" |
| }, |
| { |
| "first": "Richard", |
| "middle": [], |
| "last": "Xiao", |
| "suffix": "" |
| }, |
| { |
| "first": "Yukio", |
| "middle": [], |
| "last": "Tono", |
| "suffix": "" |
| } |
| ], |
| "year": 2006, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Tony McEnery, Richard Xiao, and Yukio Tono. 2006. L2 acquisition of grammatical morphemes. In Corpus- based language studies. An advanced resource book. Routledge.", |
| "links": null |
| }, |
| "BIBREF23": { |
| "ref_id": "b23", |
| "title": "Google books n-gram corpus used as a grammar checker", |
| "authors": [ |
| { |
| "first": "Rogelio", |
| "middle": [], |
| "last": "Nazar", |
| "suffix": "" |
| }, |
| { |
| "first": "Irene", |
| "middle": [], |
| "last": "Renau", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "EACL 2012 Proceedings of the Second Workshop on Computational Linguistics and Writing (CLW 2012)", |
| "volume": "", |
| "issue": "", |
| "pages": "27--34", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rogelio Nazar and Irene Renau. 2012. Google books n-gram corpus used as a grammar checker. In EACL 2012 Proceedings of the Second Workshop on Com- putational Linguistics and Writing (CLW 2012), pages 27-34.", |
| "links": null |
| }, |
| "BIBREF24": { |
| "ref_id": "b24", |
| "title": "Automatic error detection method for Japanese case particles in Japanese language learners' writing", |
| "authors": [ |
| { |
| "first": "Hiromi", |
| "middle": [], |
| "last": "Oyama", |
| "suffix": "" |
| }, |
| { |
| "first": "Yuji", |
| "middle": [], |
| "last": "Matsumoto", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Corpus, ICT, and Language Education", |
| "volume": "", |
| "issue": "", |
| "pages": "235--245", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hiromi Oyama and Yuji Matsumoto. 2010. Automatic error detection method for Japanese case particles in Japanese language learners' writing. In Corpus, ICT, and Language Education, pages 235-245.", |
| "links": null |
| }, |
| "BIBREF25": { |
| "ref_id": "b25", |
| "title": "Real Academia de la Lengua Espa\u00f1ola RAE", |
| "authors": [], |
| "year": 2006, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Real Academia de la Lengua Espa\u00f1ola RAE. 2006. Dic- cionario panhisp\u00e1nico de dudas. Real Santillana.", |
| "links": null |
| }, |
| "BIBREF26": { |
| "ref_id": "b26", |
| "title": "New Grammar of Spanish Language (In Spanish: Nueva gram\u00e1tica de la lengua espa\u00f1ola)", |
| "authors": [], |
| "year": 2009, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Real Academia de la Lengua Espa\u00f1ola RAE. 2009. New Grammar of Spanish Language (In Spanish: Nueva gram\u00e1tica de la lengua espa\u00f1ola). Espasa Calpe.", |
| "links": null |
| }, |
| "BIBREF27": { |
| "ref_id": "b27", |
| "title": "Annotating ESL errors: Challenges and rewards", |
| "authors": [ |
| { |
| "first": "Alla", |
| "middle": [], |
| "last": "Rozovskaya", |
| "suffix": "" |
| }, |
| { |
| "first": "Dan", |
| "middle": [], |
| "last": "Roth", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Proceedings of NAACL'10 Workshop on Innovative Use of NLP for Building Educational Applications. University of Illinois at", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Alla Rozovskaya and Dan Roth. 2010. Annotating ESL errors: Challenges and rewards. In Proceedings of NAACL'10 Workshop on Innovative Use of NLP for Building Educational Applications. University of Illi- nois at Urbana-Champ.", |
| "links": null |
| }, |
| "BIBREF28": { |
| "ref_id": "b28", |
| "title": "Native judgments of non-native usage: Experiments in preposition error detection", |
| "authors": [ |
| { |
| "first": "Joel", |
| "middle": [], |
| "last": "Tetreault", |
| "suffix": "" |
| }, |
| { |
| "first": "Martin", |
| "middle": [], |
| "last": "Chodorow", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of the Workshop on Human Judgments in Computational Linguistics at the COLING 2008", |
| "volume": "", |
| "issue": "", |
| "pages": "24--32", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Joel Tetreault and Martin Chodorow. 2008. Native judg- ments of non-native usage: Experiments in preposi- tion error detection. In Proceedings of the Workshop on Human Judgments in Computational Linguistics at the COLING 2008, pages 24-32.", |
| "links": null |
| }, |
| "BIBREF29": { |
| "ref_id": "b29", |
| "title": "Writing assistants and automatic lexical error correction", |
| "authors": [ |
| { |
| "first": "Leo", |
| "middle": [], |
| "last": "Wanner", |
| "suffix": "" |
| }, |
| { |
| "first": "Serge", |
| "middle": [], |
| "last": "Verlinde", |
| "suffix": "" |
| }, |
| { |
| "first": "Margarita", |
| "middle": [], |
| "last": "Alonso", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Proceedings of the eLex 2013 conference", |
| "volume": "", |
| "issue": "", |
| "pages": "472--487", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Leo Wanner, Serge Verlinde, and Margarita Alonso. 2013. Writing assistants and automatic lexical error correction. In Proceedings of the eLex 2013 confer- ence, pages 472-487.", |
| "links": null |
| }, |
| "BIBREF30": { |
| "ref_id": "b30", |
| "title": "A web-based English proofing system for English as a second language users", |
| "authors": [ |
| { |
| "first": "Xing", |
| "middle": [], |
| "last": "Yi", |
| "suffix": "" |
| }, |
| { |
| "first": "Jianfeng", |
| "middle": [], |
| "last": "Gao", |
| "suffix": "" |
| }, |
| { |
| "first": "William", |
| "middle": [ |
| "B" |
| ], |
| "last": "Dolan", |
| "suffix": "" |
| } |
| ], |
| "year": 2008, |
| "venue": "Proceedings of the International Joint Conference on Natural Language Processing", |
| "volume": "", |
| "issue": "", |
| "pages": "619--624", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Xing Yi, Jianfeng Gao, and William B. Dolan. 2008. A web-based English proofing system for English as a second language users. In Proceedings of the Interna- tional Joint Conference on Natural Language Process- ing, pages 619-624, Hyderabad, India.", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "TABREF0": { |
| "content": "<table><tr><td>PACLIC 28</td></tr><tr><td>2 Spanish articles</td></tr><tr><td>2.1 General overview</td></tr><tr><td>! 235</td></tr></table>", |
| "text": "Copyright 2014 by Mar \u00eda del Pilar Valverde Iban \u1ebdz and Akira Ohtani 28th Pacific Asia Conference on Language, Information and Computation pages 234-243", |
| "html": null, |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF2": { |
| "content": "<table/>", |
| "text": "", |
| "html": null, |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF3": { |
| "content": "<table><tr><td/><td>: Tags</td></tr><tr><td colspan=\"2\">Missing article (AD, AI) A missing error oc-</td></tr><tr><td colspan=\"2\">curs when the learner does not use any article but</td></tr><tr><td colspan=\"2\">the sentence should contain one: definite, as in</td></tr><tr><td colspan=\"2\">(5) (AD|AD|AD|AD||AD) 3 or indefinite as in (6)</td></tr><tr><td colspan=\"2\">(AI|NC|AI|AI||AI).</td></tr><tr><td>(5)</td><td>Originalmente el espa\u00f1ol y el portugu\u00e9s son</td></tr><tr><td/><td>categorizados en mismo grupo ling\u00fc\u00edstico, la</td></tr><tr><td/><td>lengua rom\u00e1nica.</td></tr><tr><td/><td>'Originally Spanish and Portuguese are cate-</td></tr><tr><td/><td>gorized in the same linguistic group, the ro-</td></tr><tr><td/><td>mance language.'</td></tr><tr><td>(6)</td><td>Osu est\u00e1 cerca del barrio de Sakae que es</td></tr><tr><td/><td>centro comercial muy animado y moderno.</td></tr><tr><td/><td>'Osu is near Sakae area which is</td></tr><tr><td/><td>a very lively and modern commercial district.'</td></tr><tr><td colspan=\"2\">Extraneous article (E) An extraneous article er-</td></tr><tr><td colspan=\"2\">ror occurs when the article used by the learner is not</td></tr><tr><td colspan=\"2\">necessary (zero article should be used instead), as in</td></tr><tr><td colspan=\"2\">(7) (E|E|E|E||E).</td></tr></table>", |
| "text": "El objetivo de este trabajo es conocer c\u00f3mo propag\u00f3 el tomate como la verdura comestible desde el continente americano. 'The goal of this paper is to know how tomato spreaded as an edible vegetable from the American continent.'", |
| "html": null, |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF5": { |
| "content": "<table><tr><td>NE1:# NE2: ! AD AI C E NC OK</td><td colspan=\"3\">AD AI C E NC OK Tot 2 0 1 0 10 44 31 2 0 0 0 2 9 5 1 2 2 6 34 0 23 0 0 2 10 73 4 57 0 0 0 1 0 1 0 5 1 5 7 2 119 139</td></tr><tr><td>Tot</td><td>39</td><td>8 32 68</td><td>6 147 300</td></tr></table>", |
| "text": "Confusion matrix for E1 and E2 annotators.", |
| "html": null, |
| "type_str": "table", |
| "num": null |
| }, |
| "TABREF6": { |
| "content": "<table/>", |
| "text": "", |
| "html": null, |
| "type_str": "table", |
| "num": null |
| } |
| } |
| } |
| } |
