| { | |
| "paper_id": "C10-1002", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T12:55:54.057021Z" | |
| }, | |
| "title": "Identifying Multi-word Expressions by Leveraging Morphological and Syntactic Idiosyncrasy", | |
| "authors": [ | |
| { | |
| "first": "Hassan", | |
| "middle": [], | |
| "last": "Al-Haj", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Language Technologies Institute Carnegie Mellon University", | |
| "location": {} | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "University of Haifa", | |
| "location": {} | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "Multi-word expressions constitute a significant portion of the lexicon of every natural language, and handling them correctly is mandatory for various NLP applications. Yet such entities are notoriously hard to define, and are consequently missing from standard lexicons and dictionaries. Multi-word expressions exhibit idiosyncratic behavior on various levels: orthographic, morphological, syntactic and semantic. In this work we take advantage of the morphological and syntactic idiosyncrasy of Hebrew noun compounds and employ it to extract such expressions from text corpora. We show that relying on linguistic information dramatically improves the accuracy of compound extraction, reducing over one third of the errors compared with the best baseline.", | |
| "pdf_parse": { | |
| "paper_id": "C10-1002", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "Multi-word expressions constitute a significant portion of the lexicon of every natural language, and handling them correctly is mandatory for various NLP applications. Yet such entities are notoriously hard to define, and are consequently missing from standard lexicons and dictionaries. Multi-word expressions exhibit idiosyncratic behavior on various levels: orthographic, morphological, syntactic and semantic. In this work we take advantage of the morphological and syntactic idiosyncrasy of Hebrew noun compounds and employ it to extract such expressions from text corpora. We show that relying on linguistic information dramatically improves the accuracy of compound extraction, reducing over one third of the errors compared with the best baseline.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Multi-word expressions (MWEs) are notoriously hard to define. They span a range of constructions, from completely frozen, semantically opaque idiomatic expressions, to frequent but morphologically productive and semantically compositional collocations. Various linguistic processes (orthographic, morphological, syntactic, semantic, and cognitive) apply to MWEs in idiosyncratic ways. Notably, MWEs blur the distinction between the lexicon and the grammar, since they often have some properties of words and some of phrases.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "In this work we define MWEs as expressions whose linguistic properties (morphological, syntactic or semantic) are not directly derived from the properties of their word constituents. This is a functional definition, driven by a practical motivation: any natural language processing (NLP) application that cares about morphology, syntax or semantics must consequently store MWEs in the lexicon.", | |
| "cite_spans": [ | |
| { | |
| "start": 282, | |
| "end": 287, | |
| "text": "(NLP)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "MWEs are numerous and constitute a significant portion of the lexicon of any natural language. They are a heterogeneous class of constructions with diverse sets of characteristics. Morphologically, some MWEs allow some of their constituents to freely inflect while restricting (or even preventing) the inflection of other constituents. MWEs may allow constituents to undergo non-standard morphological inflections that they would not undergo in isolation. Some MWEs contain words that never occur outside the context of the MWE. Syntactically, some MWEs appear in one rigid pattern (and a fixed order), while others permit various syntactic transformations. Semantically, the compositionality of MWEs (i.e., the degree to which the meaning of the whole expression results from combining the meanings of its individual words when they occur in isolation) is gradual.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "These morphological, syntactic and semantic idiosyncrasies make MWEs a challenge for NLP applications (Sag et al., 2002) . They are even more challenging in languages with complex morphology, because of the unique interaction of morphological and orthographic processes with the lexical specification of MWEs (Oflazer et al., 2004; Alegria et al., 2004) .", | |
| "cite_spans": [ | |
| { | |
| "start": 102, | |
| "end": 120, | |
| "text": "(Sag et al., 2002)", | |
| "ref_id": "BIBREF22" | |
| }, | |
| { | |
| "start": 309, | |
| "end": 331, | |
| "text": "(Oflazer et al., 2004;", | |
| "ref_id": "BIBREF18" | |
| }, | |
| { | |
| "start": 332, | |
| "end": 353, | |
| "text": "Alegria et al., 2004)", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Because the idiosyncratic features of MWEs cannot be predicted on the basis of their component words, they must be stored in the lexicon of NLP applications. Handling MWEs correctly is beneficial for a variety of applications, including information retrieval, building ontologies, text alignment, and machine translation. Automatic identification and corpus-based extraction of MWEs is thus crucial for such (and several other) applications.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "In this work we describe an approach that leverages the morphological and syntactic idiosyncrasy of a certain class of Hebrew 1 MWEs, namely noun compounds, to help identify such expressions in texts. While the main contribution of this work is a system that can distinguish between MWE and non-MWE instances of a particular construction in Hebrew, thereby facilitating faster and more accurate integration of MWEs in a large-coverage lexicon of the language, we believe that it carries added value to anyone interested in MWEs. The technique that we propose here should be applicable in principle to any language in which MWEs exhibit linguistically idiosyncratic behavior.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "We describe the properties of Hebrew nounnoun constructions in Section 2, and specify the irregularities exhibited by compounds. Section 3 presents the experimental setup and the main results. Compared with the best (collocation-based) baseline, our approach reduces over 30% of the errors, yielding accuracy of over 80%. We discuss related work in Section 4 and conclude with suggestions for future research.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "We focus on Hebrew noun-noun constructions; these are extremely frequent constructions, and while many of them are fully compositional, others, called noun compounds (or just compounds) here, are clearly MWEs. We first discuss the general construction and then describe the peculiar, idiosyncratic properties of compounds.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Hebrew noun-noun constructions", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Hebrew nouns inflect for number (singular and plural) and, when the noun denotes an animate entity, for gender (masculine and feminine). In addition, nouns come in three states: indefinite, definite and a construct state that is used in genitive constructions. Table 1 demonstrates the paradigm. A noun-noun construction (henceforth NNC) consists of a construct-state noun, called head here, followed by a noun phrase, the modifier (Borer, 1988; Borer, 1996; Glinert, 1989) .", | |
| "cite_spans": [ | |
| { | |
| "start": 432, | |
| "end": 445, | |
| "text": "(Borer, 1988;", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 446, | |
| "end": 458, | |
| "text": "Borer, 1996;", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 459, | |
| "end": 473, | |
| "text": "Glinert, 1989)", | |
| "ref_id": "BIBREF13" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 261, | |
| "end": 295, | |
| "text": "Table 1 demonstrates the paradigm.", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "The general case", | |
| "sec_num": "2.1" | |
| }, | |
| { | |
| "text": "M/Sg F/Sg M/Pl F/Pl indefinite ild ildh ildim ildwt definite hild hildh hildim hildwt construct ild ildt ildi ildwt Table 1 : The noun paradigm, demonstrated on ild \"child\"", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 116, | |
| "end": 123, | |
| "text": "Table 1", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "State", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The semantic relation between the two is usually, but not always, related to possession (Levi, 1976) . Construct-state nouns only occur in the context of NNC, and can never occur in isolation. When a NNC is definite, the definite article is expressed on its modifier (Wintner, 2000) .", | |
| "cite_spans": [ | |
| { | |
| "start": 88, | |
| "end": 100, | |
| "text": "(Levi, 1976)", | |
| "ref_id": "BIBREF16" | |
| }, | |
| { | |
| "start": 267, | |
| "end": 282, | |
| "text": "(Wintner, 2000)", | |
| "ref_id": "BIBREF28" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "State", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "In the examples below, we explicitly indicate construct-state nouns by the morpheme '.CONST' in the gloss; and definite nouns are indicated by the morpheme 'the-'. We provide both a literal and a non-literal meaning of the MWE examples. Expressions that have a literal, but not the expected MWE meaning, are preceded by '#'.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "State", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Example 1 (Noun-noun constructions) hxlTt hw'dh decision.CONST the-committee \"the committee decision\" 'wrk h'itwn editor.CONST the-journal \"the journal editor\" 'wrk din editor.CONST law \"law editor\" =\u21d2 lawyer bti xwlim houses.CONST patients \"patient houses\" =\u21d2 hospitals 2.2 Noun compounds: Linguistic properties While many of the NNCs are free, compositional combinations of words, some are not; we use the term noun compounds for the latter group. Compounds typically (but not necessarily) have noncompositional meaning; presumably due to their opaque, more lexical meaning, they also differ from other NNCs in their morphological and syntactic behavior. Some of these distinctive properties are listed below, to motivate the methodology that we propose in Section 3 to distinguish between compounds and non-MWE NNCs.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "State", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "When a NNC consists of two nouns, the second can typically occur in either singular or plural form. Compounds often limit the possibilities to only one of those.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Limited inflection", | |
| "sec_num": "2.2.1" | |
| }, | |
| { | |
| "text": "'wrki h'itwnim editors-.CONST the-journals \"the journals' editors\"", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example 2 (No plural form of the modifier)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "'wrki hdin editors.CONST the-law \"the law editors\" =\u21d2 the lawyers #wrki hdinim editors.CONST the-laws Example 3 (No singular form of the modifier) kiwwn hrwx direction.CONST the-wind \"the wind's direction\" kiwwn hrwxwt direction.CONST the-winds \"the winds' direction\" sw\u0161nt h-rwxwt lily.CONST the-winds \"lily of the winds\" =\u21d2 compass rose #\u0161w\u0161nt h-rwx lily.CONST the-wind", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example 2 (No plural form of the modifier)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Since NNCs typically denote genitive (possessive) constructions, they can be paraphrased by a construction that uses the genitive preposition\u0161l \"of\" (or, in some cases, other prepositions). These syntactic variants are often restricted in the case of compounds.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Limited syntactic variation", | |
| "sec_num": "2.2.2" | |
| }, | |
| { | |
| "text": "h'wrk\u0161l h'itwn the-editor of the-journal \"the journal editor\" #h'wrk\u0161l hdin the-editor of the-law Example 5 (Limited paraphrasing) m'il cmr coat.CONST wool \"wool coat\" m'il mcmr coat from-wool \"wool coat\" cmr pldh wool.CONST steel \"steel wool\" =\u21d2 steel wool #cmr mpldh wool from-steel", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example 4 (Limited paraphrasing)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "NNCs typically allow adjectival modification of either of their constituents. Since compounds tend to be more semantically opaque, it is often only possible to modify the entire compound, but not any of the constituents. In the following example, note that 'wrkt \"editor\" is feminine, whereas 'itwn \"journal\" is masculine; adjectives must agree on gender with the noun they modify.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Limited syntactic modification", | |
| "sec_num": "2.2.3" | |
| }, | |
| { | |
| "text": "'wrkt h'itwn editor-f.CONST the-journal-m \"the journal editor\" 'wrkt h'itwn hxd\u0161h editor-f.CONST the-journal-m the-new-f \"the new editor of the journal\" 'wrkt h'itwn hxd\u0161 editor-f.CONST the-journal-m the-new-m \"the editor of the new journal\" 'wrkt hdin hxd\u0161h editor-f.CONST the-law-m the-new-f \"the new law editor\" =\u21d2 the new lawyer #'wrkt hdin hxd\u0161 editor-f.CONST the-law-m the-new-m", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Example 6 (Limited adjectival modification)", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Two NNCs that share a common head can be conjoined using the coordinating conjunction w \"and\". This possibility is often blocked in the case of compounds. In this section we describe a system that identifies noun compounds in Hebrew text, and extracts them in order to extend the lexicon. We capitalize on the morphological and syntactic irregularities of noun compounds described in Section 2.2.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Limited coordination", | |
| "sec_num": "2.2.4" | |
| }, | |
| { | |
| "text": "Given a large monolingual corpus, the text is first morphologically analyzed and disambiguated. Then, all NNCs (candidate noun compounds) are extracted from the morphologically disambiguated text. For each candidate noun compound we define a set of features (Section 3.3) based on the idiosyncratic morphological and syntactic properties defined in Section 2.2. These features inform a support vector machine classifier which is then used to identify the noun compounds in the set of NNCs with high accuracy (Section 3.5).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Limited coordination", | |
| "sec_num": "2.2.4" | |
| }, | |
| { | |
| "text": "We use (a subset of) the Corpus of Contemporary Hebrew (Itai and Wintner, 2008) The entire corpus was morphologically analyzed (Yona and Wintner, 2008; Itai and Wintner, 2008) and POS-tagged (Bar-haim et al., 2008) ; note that no syntactic parser is available for Hebrew. From the morphologically disambiguated corpus, we extract all bi-grams in which the first token is a noun in the construct state and the second token is a noun that is not in the construct state, i.e., all two-word NNC candidates.", | |
| "cite_spans": [ | |
| { | |
| "start": 55, | |
| "end": 79, | |
| "text": "(Itai and Wintner, 2008)", | |
| "ref_id": "BIBREF14" | |
| }, | |
| { | |
| "start": 127, | |
| "end": 151, | |
| "text": "(Yona and Wintner, 2008;", | |
| "ref_id": "BIBREF29" | |
| }, | |
| { | |
| "start": 152, | |
| "end": 175, | |
| "text": "Itai and Wintner, 2008)", | |
| "ref_id": "BIBREF14" | |
| }, | |
| { | |
| "start": 180, | |
| "end": 214, | |
| "text": "POS-tagged (Bar-haim et al., 2008)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Resources", | |
| "sec_num": "3.1" | |
| }, | |
| { | |
| "text": "For training and evaluation, we select the NNCs that occur at least 100 times in the corpus, yielding 1060 NNCs. These NNCs were annotated by three annotators, who were asked to classify them to the following four groups: compounds (+); non-compounds (-); unsure (0); and errors of the morphological processor (i.e., the candidate is not a NNC at all). We adopt a conservative approach in combining the three annotations. First, we eliminate 204 NNCs that were tagged as errors by at least one annotator. For the remaining NNCs, a candidate is considered a compound or a non-compound only if all three annotators agree on its classification. This reduces the annotated data to 463 instances, of which 205 are compounds and 258 are clear cases of non-compound NNCs. 2", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Annotation", | |
| "sec_num": "3.2" | |
| }, | |
| { | |
| "text": "We define a set of features based on the idiosyncratic properties of noun compounds defined in Section 2.2. For each candidate NNC, we compute counts which reflect the likelihood of it exhibiting one of the linguistic properties.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "Refer back to Section 2.2. We focus on the property of limited inflection (Section 2.2.1), and define features 1-8 to reflect it. To reflect limited syntactic variation (Section 2.2.2) we define features 9-10. Feature 11 addresses the phenomenon of limited coordination (Section 2.2.4). To reflect limited syntactic modification (Section 2.2.3) we define feature 12. .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "For each NNC candidate N 1 N 2 , the following features are defined:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "1. The number of occurrences of the NNC in which both constituents are in singular.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "2. The number of occurrences of the NNC in which N 1 is in singular and N 2 is in plural.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "3. The number of occurrences of the NNC in which N 1 is in plural and N 2 is in singular.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "4. The number of occurrences of the NNC in which both constituents are in plural.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "5. The number of occurrences of N 1 in plural outside the expression.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "6. The number of occurrences of N 1 in singular outside the expression.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "7. The number of occurrences of N 2 in plural outside the expression.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "8. The number of occurrences of N 2 in singular outside the expression.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "9. The number of occurrences of N 1\u0161 l N 2 \"N 1 of N 2 \" in the corpus.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "10. The number of occurrences of N 1 m N 2 \"N 1 from N 2 \" in the corpus.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "11. The number of occurrences of N 1 N 2 w N 3 \"N 1 N 2 and N 3 \" in the corpus, where N 3 is an indefinite, non-construct-state noun.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "12. The number of occurrences of N 1 N 2 Adj in the corpus, where the adjective Adj agrees with N 2 on both gender and number, while disagreeing with N 1 on at least one of these attributes.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "We also define four features that represent known collocation measures (Evert and Krenn, 2001) : Point-wise mutual information (PMI); T-Score; log-likelihood; and the raw frequency of N 1 N 2 in the corpus. 3", | |
| "cite_spans": [ | |
| { | |
| "start": 71, | |
| "end": 94, | |
| "text": "(Evert and Krenn, 2001)", | |
| "ref_id": "BIBREF10" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "3 A detailed description of these measures is given by Manning and Sch\u00fctze (1999, Chapter 5) ; see also http: //www.collocations.de/, where several other association measures are discussed as well.", | |
| "cite_spans": [ | |
| { | |
| "start": 55, | |
| "end": 92, | |
| "text": "Manning and Sch\u00fctze (1999, Chapter 5)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Linguistically-motivated features", | |
| "sec_num": "3.3" | |
| }, | |
| { | |
| "text": "For each NNC in the annotated set of Section 3.2 we create a vector of the 16 features described in Section 3.3 (12 linguistically-motivated features plus four collocation measures). We obtain a list of 463 instances, of which 205 are positive examples (noun compounds) and 258 are negative. We use this set for training and evaluation of a two class soft margin SVM classifier (Chang and Lin, 2001 ) with a radial basis function kernel. We experiment below with different combinations of features, where for each combination we use 10fold cross-validation over the 463 NNcs to evaluate the classifier. We report Precision, Recall, Fscore and Accuracy (averaged over the 10 folds).", | |
| "cite_spans": [ | |
| { | |
| "start": 378, | |
| "end": 398, | |
| "text": "(Chang and Lin, 2001", | |
| "ref_id": "BIBREF8" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Training and evaluation", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "The results of the different classifiers that we trained are given in Table 4 . The first four rows of the table show the performance of classifiers trained using each of the four different collocation measure features alone. Both PMI and Loglikelihood outperform the other collocation measures, with an F-score of 60, which we consider our baseline. We also report the performance of two combinations of collocation measures, which yield small improvement. The best combinations provide accuracy of about 70% and F-score of 63.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 70, | |
| "end": 77, | |
| "text": "Table 4", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Results", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "The remaining rows report results using the linguistically-motivated features (LMF) of Section 3.3. These features alone yield accuracy of 77.75% and an F-score of 76. Adding also Loglikelihood improves F-score by 1.16 and accuracy by 1.29%. Finally, using Log-likelihood with a subset of the LMF consisting of features 1-2, 4-6, 9-10 and 12 (see below) yields the best results, namely accuracy of over 80% and F-score of 78.85, reflecting a reduction of over one third in classification error rate compared with the baseline.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Results", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "We search for the combination of linguisticallymotivated features that would yield the best performance. Training a classifier on all possible feature combinations is clearly infeasible. Instead, we follow a more efficient greedy approach, whereby we start with the best collocation mea- Table 4 : Results: 10-Fold accuracy, precision, recall, and F-score for classifiers trained using different combinations of features. LMF stands for linguistically-motivated features sure, Log-likelihood, and add other features one at a time, in the order in which they are listed in Section 3.3. After adding each feature the classifier is retrained; the feature is retained in the feature set only if adding it improves the 10-fold F-score of the current feature set. Table 5 lists the results of this experiment. For each feature set the difference in the 10-fold Fscore compared to the previous feature set is listed in parentheses. The results show that the best feature combination improves the F-score by 1.26, compared with using all features. This experiments shows that features 3, 7, 8 and 11 turn out not to be useful, and the classifier is more accurate without them. We also tried this approach with PMI as the starting feature, with very similar results. 66.66 (+1.78) Log-likelihood,1-2,4-5 70.00 (+3.33) Log-likelihood,1-2,4-6 74.37 (+4.37) Log-likelihood,1-2,4-7 73.78 (-0.58) Log-likelihood,1-2,4-6,8 73.58 (-0.79) Log-likelihood,1-2,4-6,9", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 288, | |
| "end": 295, | |
| "text": "Table 4", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 758, | |
| "end": 765, | |
| "text": "Table 5", | |
| "ref_id": "TABREF7" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Optimizing feature combination", | |
| "sec_num": "3.6" | |
| }, | |
| { | |
| "text": "78.72 (+4.35) Log-likelihood,1-2,4-6,9-10 78.83 (+0.10) Log-likelihood,1-2,4-6,9-11 77.37 (-1.46) Log-likelihood,1-2,4-6,9-10,12 78.85 (+0.02) ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Optimizing feature combination", | |
| "sec_num": "3.6" | |
| }, | |
| { | |
| "text": "There has been a growing awareness in the research community of the problems that MWEs pose, both in linguistics and in NLP (Villavicencio et al., 2005) . Recent works address the definition, lexical representation and computational processing of MWEs, as well as algorithms for extracting them from data.", | |
| "cite_spans": [ | |
| { | |
| "start": 124, | |
| "end": 152, | |
| "text": "(Villavicencio et al., 2005)", | |
| "ref_id": "BIBREF27" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "Focusing on acquisition of MWEs, early approaches concentrated on their collocational behavior (Church and Hanks, 1989) . Pecina (2008) compares 55 different association measures in ranking German Adj-N and PP-Verb collocation candidates. This work shows that combining different collocation measures using standard statistical-classification methods (such as Linear Logistic Regression and Neural Networks) gives a significant improvement over using a single collocation measure. Our results show that this is indeed the case, but the contribution of collocation methods is limited, and more information is needed in order to distinguish frequent collocations from bona fide MWEs.", | |
| "cite_spans": [ | |
| { | |
| "start": 95, | |
| "end": 119, | |
| "text": "(Church and Hanks, 1989)", | |
| "ref_id": "BIBREF9" | |
| }, | |
| { | |
| "start": 122, | |
| "end": 135, | |
| "text": "Pecina (2008)", | |
| "ref_id": "BIBREF20" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "Other works show that adding linguistic information to collocation measures can improve identification accuracy. Several approaches rely on the semantic opacity of MWEs; but very few semantic resources are available for Hebrew (the Hebrew WordNet (Ordan and Wintner, 2007) , the only lexical semantic resource for this language, is small and too limited). Instead, we capital-ize on the morphological and syntactic irregularities that MWEs exhibit, using computational resources that are more readily-available. Ramisch et al. (2008) evaluate a number of association measures on the task of identifying English Verb-Particle Constructions and German Adjective-Noun pairs. They show that adding linguistic information (mostly POS and POSsequence patterns) to the association measure yields a significant improvement in performance over using pure frequency. We follow this line of research by defining a number of syntactic patterns as a source of linguistic information. In addition, our linguistic features are much more specific to the phenomenon we are interested in, and the syntactic patterns are enriched by morphological information pertaining to the idiosyncrasy of MWEs; we believe that this explains the improved performance compared to the baseline.", | |
| "cite_spans": [ | |
| { | |
| "start": 247, | |
| "end": 272, | |
| "text": "(Ordan and Wintner, 2007)", | |
| "ref_id": "BIBREF19" | |
| }, | |
| { | |
| "start": 512, | |
| "end": 533, | |
| "text": "Ramisch et al. (2008)", | |
| "ref_id": "BIBREF21" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "Several works address the lexical fixedness or syntactic fixedness of (certain types of) MWEs in order to extract them from texts. An expression is considered lexically fixed if replacing any of its constituents by a semantically (and syntactically) similar word generally results in an invalid or literal expression. Syntactically fixed expressions prohibit (or restrict) syntactic variation.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "For example, Van de Cruys and Villada Moir\u00f3n (2007) use lexical fixedness to extract Dutch Verb-Noun idiomatic combinations (VNICs). Bannard (2007) uses syntactic fixedness to identify English VNICs. Another work uses both the syntactic and the lexical fixedness of VNICs in order to distinguish them from non-idiomatic ones, and eventually to extract them from corpora (Fazly and Stevenson, 2006) . While these approaches are in line with ours, they require lexical semantic resources (e.g., a database that determines semantic similarity among words) and syntactic resources (parsers) that are unavailable for Hebrew (and many other languages). Our approach only requires morphological processing, which is more readily-available for several languages.", | |
| "cite_spans": [ | |
| { | |
| "start": 133, | |
| "end": 147, | |
| "text": "Bannard (2007)", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 370, | |
| "end": 397, | |
| "text": "(Fazly and Stevenson, 2006)", | |
| "ref_id": "BIBREF12" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "Another unique feature of our work is that it computationally addresses Hebrew (and, more generally, Semitic) MWEs for the first time. Berman and Ravid (1986) define the dictionary degree of noun compounds in Hebrew as their closeness to a single word from a grammatical point of view, as judged by the manner in which they are grasped by language speakers. A group of 120 Hebrew speakers were asked to assign a dictionary degree (from 1 to 5) to a list of 30 noun compounds. An analysis of the questionnaire results revealed that language speaker share a common dictionary, where the highest degree of agreement was achieved on the ends of the dictionary degree spectrum. Another conclusion is that both the pragmatic uses of the noun compound and the semantic relation between its constituents define the dictionary degree of the compound. Not having access to semantic and pragmatic knowledge, we are trying to approximate it using morphology. Attia (2005) proposes methods to process fixed, semi-fixed, and syntactically-flexible Arabic MWEs (adopting the classification and the terminology of Sag et al. (2002) ). Fabri (2009) provides an overview of the different types of compounds (14 in total) in present-day Maltese, focusing on one type of compounds consisting of an adjective followed by a noun. He also provides morphological, syntactic, and semantic properties of this group which distinguishes them from other non-compound constructions. Automatic identification of MWEs is not addressed in either of these works.", | |
| "cite_spans": [ | |
| { | |
| "start": 135, | |
| "end": 158, | |
| "text": "Berman and Ravid (1986)", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 947, | |
| "end": 959, | |
| "text": "Attia (2005)", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 1098, | |
| "end": 1115, | |
| "text": "Sag et al. (2002)", | |
| "ref_id": "BIBREF22" | |
| }, | |
| { | |
| "start": 1119, | |
| "end": 1131, | |
| "text": "Fabri (2009)", | |
| "ref_id": "BIBREF11" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Related work", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "We described a system that can identify Hebrew noun compounds with high accuracy, distinguishing them from non-idiomatic noun-noun constructions. The methodology we advocate is based on careful examination of the linguistic peculiarities of the construction, followed by corpus-based approximation of these properties via a general machine learning algorithm that is fed with features based on the linguistic properties. While our application is limited to a particular construction in a particular language, we are confident that it can be equally well applied to other constructions and other languages, as long as the targeted MWEs exhibit a consistent set of irregular features (especially in the morphology).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and future work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "This work can be extended in various directions. Addressing other constructions is relatively easy, and requires only a theoretical linguistic investigation of the construction. We are currently interested in extending the system to cope also with Adjective-Noun, Noun-Adjective and Verb-Preposition constructions in Hebrew.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and future work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "The accuracy of MWE acquisition systems can be further improved by combining our morphological and syntactic features with semantically informed features such as translational entropy computed from a parallel corpus (Villada Moir\u00f3n and Tiedemann, 2006) , or features that can capture the local linguistic context of the expression using latent semantic analysis (Katz and Giesbrecht, 2006) . We are currently working on the former direction (Tsvetkov and Wintner, 2010b) , utilizing a small Hebrew-English parallel corpus (Tsvetkov and Wintner, 2010a) .", | |
| "cite_spans": [ | |
| { | |
| "start": 216, | |
| "end": 252, | |
| "text": "(Villada Moir\u00f3n and Tiedemann, 2006)", | |
| "ref_id": "BIBREF26" | |
| }, | |
| { | |
| "start": 362, | |
| "end": 389, | |
| "text": "(Katz and Giesbrecht, 2006)", | |
| "ref_id": "BIBREF15" | |
| }, | |
| { | |
| "start": 441, | |
| "end": 470, | |
| "text": "(Tsvetkov and Wintner, 2010b)", | |
| "ref_id": "BIBREF24" | |
| }, | |
| { | |
| "start": 522, | |
| "end": 551, | |
| "text": "(Tsvetkov and Wintner, 2010a)", | |
| "ref_id": "BIBREF23" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and future work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "Finally, we are interested in evaluating the methodology proposed in this paper to other languages with complex morphology, in particular to Arabic. We leave this direction to future research.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and future work", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "To facilitate readability we use a transliteration of Hebrew using Roman characters; the letters used, in Hebrew lexicographic order, are abgdhwzxTiklmns'pcqr\u0161t.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "This annotated corpus is freely available for download.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "This research was supported by THE IS-RAEL SCIENCE FOUNDATION (grants No. 137/06, 1269/07). We are grateful to Alon Itai for his continuous help and advice throughout the course of this project, and to Bracha Nir for very useful comments. We also wish to thank Yulia Tsvetkov and Gily Chen for their annotation work.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Acknowledgments", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Representation and treatment of multiword expressions in Basque", | |
| "authors": [ | |
| { | |
| "first": "I\u00f1aki", | |
| "middle": [], | |
| "last": "Alegria", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Olatz", | |
| "middle": [], | |
| "last": "Ansa", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xabier", | |
| "middle": [], | |
| "last": "Artola", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nerea", | |
| "middle": [], | |
| "last": "Ezeiza", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Koldo", | |
| "middle": [], | |
| "last": "Gojenola", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ruben", | |
| "middle": [], | |
| "last": "Urizar", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Second ACL Workshop on Multiword Expressions: Integrating Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "48--55", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Alegria, I\u00f1aki, Olatz Ansa, Xabier Artola, Nerea Ezeiza, Koldo Gojenola, and Ruben Urizar. 2004. Representation and treatment of multiword expres- sions in Basque. In Tanaka, Takaaki, Aline Villavi- cencio, Francis Bond, and Anna Korhonen, editors, Second ACL Workshop on Multiword Expressions: Integrating Processing, pages 48-55, Barcelona, Spain, July. Association for Computational Linguis- tics.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Mohammed A. Attia The University of Manchester School of Informatics mohammed", | |
| "authors": [ | |
| { | |
| "first": "Mohammed", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Attia", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Attia, Mohammed A. 2005. Accommodat- ing multiword expressions in an lfg grammar. The ParGram Meeting, Japan September 2005, September. Mohammed A. Attia The Univer- sity of Manchester School of Informatics mo- [email protected].", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "A measure of syntactic flexibility for automatically identifying multiword expressions in corpora", | |
| "authors": [ | |
| { | |
| "first": "Colin", | |
| "middle": [], | |
| "last": "Bannard", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "Proceedings of the Workshop on A Broader Perspective on Multiword Expressions", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--8", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Bannard, Colin. 2007. A measure of syntactic flexibil- ity for automatically identifying multiword expres- sions in corpora. In Proceedings of the Workshop on A Broader Perspective on Multiword Expressions, pages 1-8. Association for Computational Linguis- tics.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Part-of-speech tagging of Modern Hebrew text", | |
| "authors": [ | |
| { | |
| "first": "Roy", | |
| "middle": [], | |
| "last": "Bar-Haim", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Khalil", | |
| "middle": [], | |
| "last": "Sima'an", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yoad", | |
| "middle": [], | |
| "last": "Winter", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Natural Language Engineering", | |
| "volume": "14", | |
| "issue": "2", | |
| "pages": "223--251", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Bar-haim, Roy, Khalil Sima'an, and Yoad Winter. 2008. Part-of-speech tagging of Modern Hebrew text. Natural Language Engineering, 14(2):223- 251.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Lexicalization of noun compounds", | |
| "authors": [ | |
| { | |
| "first": "Ruth", | |
| "middle": [ | |
| "A" | |
| ], | |
| "last": "Berman", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dorit", | |
| "middle": [], | |
| "last": "Ravid", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "Hebrew Linguistics", | |
| "volume": "24", | |
| "issue": "", | |
| "pages": "5--22", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Berman, Ruth A. and Dorit Ravid. 1986. Lexicaliza- tion of noun compounds. Hebrew Linguistics, 24:5- 22. In Hebrew.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "On the morphological parallelism between compounds and constructs", | |
| "authors": [ | |
| { | |
| "first": "Hagit", | |
| "middle": [], | |
| "last": "Borer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "Yearbook of Morphology 1", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "45--65", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Borer, Hagit. 1988. On the morphological parallelism between compounds and constructs. In Booij, Geert and Jaap van Marle, editors, Yearbook of Morphol- ogy 1, pages 45-65. Foris publications, Dordrecht, Holland.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "The construct in review", | |
| "authors": [ | |
| { | |
| "first": "Hagit", | |
| "middle": [], | |
| "last": "Borer", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1996, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Borer, Hagit. 1996. The construct in review.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Studies in Afroasiatic Grammar", | |
| "authors": [], | |
| "year": null, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "30--61", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "In Lecarme, Jacqueline, Jean Lowenstamm, and Ur Shlonsky, editors, Studies in Afroasiatic Gram- mar, pages 30-61. Holland Academic Graphics, The Hague.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "LIB-SVM: a library for support vector machines", | |
| "authors": [ | |
| { | |
| "first": "Chih-Chung", | |
| "middle": [], | |
| "last": "Chang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chih-Jen", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Chang, Chih-Chung and Chih-Jen Lin, 2001. LIB- SVM: a library for support vector machines. Software available at http://www.csie.ntu. edu.tw/\u02dccjlin/libsvm.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Word association norms, mutual information and lexicography (rev)", | |
| "authors": [ | |
| { | |
| "first": "Kenneth", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Church", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Patrick", | |
| "middle": [], | |
| "last": "Hanks", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "Computational Linguistics", | |
| "volume": "19", | |
| "issue": "1", | |
| "pages": "22--29", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Church, Kenneth. W. and Patrick Hanks. 1989. Word association norms, mutual information and lexicog- raphy (rev). Computational Linguistics, 19(1):22- 29.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "Methods for the qualitative evaluation of lexical association measures", | |
| "authors": [ | |
| { | |
| "first": "Stefan", | |
| "middle": [], | |
| "last": "Evert", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Brigitte", | |
| "middle": [], | |
| "last": "Krenn", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2001, | |
| "venue": "Proceedings of the 39th Annual Meeting of the Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "188--195", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Evert, Stefan and Brigitte Krenn. 2001. Methods for the qualitative evaluation of lexical association mea- sures. In Proceedings of the 39th Annual Meeting of the Association for Computational Linguistics, pages 188-195, Morristown, NJ, USA. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Compounding and adjective-noun compounds in Maltese", | |
| "authors": [ | |
| { | |
| "first": "Ray", | |
| "middle": [], | |
| "last": "Fabri", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Introducing Maltese Linguistics", | |
| "volume": "113", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Fabri, Ray. 2009. Compounding and adjective-noun compounds in Maltese. In Comrie, Bernard, Ray Fabri, Elizabeth Hume, Manwel Mifsud, Thomas Stolz, and Martine Vanhove, editors, Introducing Maltese Linguistics, volume 113 of Studies in Lan- guage Companion Series. John Benjamins.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "Automatically constructing a lexicon of verb phrase idiomatic combinations", | |
| "authors": [ | |
| { | |
| "first": "Afsaneh", | |
| "middle": [], | |
| "last": "Fazly", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Suzanne", | |
| "middle": [], | |
| "last": "Stevenson", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of the 11th Conference of the European Chapter of the Association for Computational Linguistics (EACL)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "337--344", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Fazly, Afsaneh and Suzanne Stevenson. 2006. Auto- matically constructing a lexicon of verb phrase id- iomatic combinations. In Proceedings of the 11th Conference of the European Chapter of the Associ- ation for Computational Linguistics (EACL), pages 337-344.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "The Grammar of Modern Hebrew", | |
| "authors": [ | |
| { | |
| "first": "Lewis", | |
| "middle": [], | |
| "last": "Glinert", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Glinert, Lewis. 1989. The Grammar of Modern He- brew. Cambridge University Press, Cambridge.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Language resources for Hebrew. Language Resources and Evaluation", | |
| "authors": [ | |
| { | |
| "first": "Alon", | |
| "middle": [], | |
| "last": "Itai", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "", | |
| "volume": "42", | |
| "issue": "", | |
| "pages": "75--98", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Itai, Alon and Shuly Wintner. 2008. Language re- sources for Hebrew. Language Resources and Eval- uation, 42:75-98, March.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Automatic identification of non-compositional multiword expressions using latent semantic analysis", | |
| "authors": [ | |
| { | |
| "first": "Graham", | |
| "middle": [], | |
| "last": "Katz", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Eugenie", | |
| "middle": [], | |
| "last": "Giesbrecht", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of the Workshop on Multiword Expressions: Identifying and Exploiting Underlying Properties", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "12--19", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Katz, Graham and Eugenie Giesbrecht. 2006. Au- tomatic identification of non-compositional multi- word expressions using latent semantic analysis. In Proceedings of the Workshop on Multiword Expres- sions: Identifying and Exploiting Underlying Prop- erties, pages 12-19, Sydney, Australia, July. Asso- ciation for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "A semantic analysis of Hebrew compound nominals", | |
| "authors": [ | |
| { | |
| "first": "Judith", | |
| "middle": [ | |
| "N" | |
| ], | |
| "last": "Levi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1976, | |
| "venue": "Modern Hebrew Syntax and Semantics, number 32 in North-Holland Linguistic Series", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "9--55", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Levi, Judith N. 1976. A semantic analysis of Hebrew compound nominals. In Cole, Peter, editor, Stud- ies in Modern Hebrew Syntax and Semantics, num- ber 32 in North-Holland Linguistic Series, pages 9- 55. North-Holland, Amsterdam.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Foundations of statistical natural language processing", | |
| "authors": [ | |
| { | |
| "first": "Christopher", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Manning", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Hinrich", | |
| "middle": [], | |
| "last": "Sch\u00fctze", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1999, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Manning, Christopher D. and Hinrich Sch\u00fctze. 1999. Foundations of statistical natural language process- ing. The MIT Press, Cambridge, Mass.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Integrating morphology with multi-word expression processing in Turkish", | |
| "authors": [ | |
| { | |
| "first": "Kemal", | |
| "middle": [], | |
| "last": "Oflazer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "\u00c7", | |
| "middle": [], | |
| "last": "\u00d6zlem", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bilge", | |
| "middle": [], | |
| "last": "Say", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "Second ACL Workshop on Multiword Expressions: Integrating Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "64--71", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Oflazer, Kemal,\u00d6zlem \u00c7 etinoglu, and Bilge Say. 2004. Integrating morphology with multi-word ex- pression processing in Turkish. In Tanaka, Takaaki, Aline Villavicencio, Francis Bond, and Anna Ko- rhonen, editors, Second ACL Workshop on Multi- word Expressions: Integrating Processing, pages 64-71, Barcelona, Spain, July. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Hebrew WordNet: a test case of aligning lexical databases across languages", | |
| "authors": [ | |
| { | |
| "first": "Noam", | |
| "middle": [], | |
| "last": "Ordan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "International Journal of Translation, special issue on Lexical Resources for Machine Translation", | |
| "volume": "19", | |
| "issue": "1", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ordan, Noam and Shuly Wintner. 2007. Hebrew WordNet: a test case of aligning lexical databases across languages. International Journal of Transla- tion, special issue on Lexical Resources for Machine Translation, 19(1).", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "A machine learning approach to multiword expression extraction", | |
| "authors": [ | |
| { | |
| "first": "Pavel", | |
| "middle": [], | |
| "last": "Pecina", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Proceedings of the LREC Workshop Towards a Shared Task for Multiword Expressions", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Pecina, Pavel. 2008. A machine learning approach to multiword expression extraction. In Proceedings of the LREC Workshop Towards a Shared Task for Multiword Expressions.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "An evaluation of methods for the extraction of multiword expressions", | |
| "authors": [ | |
| { | |
| "first": "Carlos", | |
| "middle": [], | |
| "last": "Ramisch", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Paulo", | |
| "middle": [], | |
| "last": "Schreiner", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Marco", | |
| "middle": [], | |
| "last": "Idiart", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alline", | |
| "middle": [], | |
| "last": "Villavicencio", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Proceedings of the LREC Workshop Towards a Shared Task for Multiword Expressions", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ramisch, Carlos, Paulo Schreiner, Marco Idiart, and Alline Villavicencio. 2008. An evaluation of meth- ods for the extraction of multiword expressions. In Proceedings of the LREC Workshop Towards a Shared Task for Multiword Expressions.", | |
| "links": null | |
| }, | |
| "BIBREF22": { | |
| "ref_id": "b22", | |
| "title": "Multiword expressions: A pain in the neck for NLP", | |
| "authors": [ | |
| { | |
| "first": "Ivan", | |
| "middle": [], | |
| "last": "Sag", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Timothy", | |
| "middle": [], | |
| "last": "Baldwin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Francis", | |
| "middle": [], | |
| "last": "Bond", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ann", | |
| "middle": [], | |
| "last": "Copestake", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dan", | |
| "middle": [], | |
| "last": "Flickinger", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "Proceedings of the Third International Conference on Intelligent Text Processing and Computational Linguistics (CICLING 2002)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1--15", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sag, Ivan, Timothy Baldwin, Francis Bond, Ann Copestake, and Dan Flickinger. 2002. Multiword expressions: A pain in the neck for NLP. In Pro- ceedings of the Third International Conference on Intelligent Text Processing and Computational Lin- guistics (CICLING 2002), pages 1-15, Mexico City, Mexico.", | |
| "links": null | |
| }, | |
| "BIBREF23": { | |
| "ref_id": "b23", | |
| "title": "Automatic acquisition of parallel corpora from websites with dynamic content", | |
| "authors": [ | |
| { | |
| "first": "Yulia", | |
| "middle": [], | |
| "last": "Tsvetkov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of the Seventh conference on International Language Resources and Evaluation (LREC'10)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "3389--3392", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tsvetkov, Yulia and Shuly Wintner. 2010a. Automatic acquisition of parallel corpora from websites with dynamic content. In Proceedings of the Seventh conference on International Language Resources and Evaluation (LREC'10), pages 3389-3392. Eu- ropean Language Resources Association (ELRA), May.", | |
| "links": null | |
| }, | |
| "BIBREF24": { | |
| "ref_id": "b24", | |
| "title": "Extraction of multi-word expressions from small parallel corpora", | |
| "authors": [ | |
| { | |
| "first": "Yulia", | |
| "middle": [], | |
| "last": "Tsvetkov", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "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": "Tsvetkov, Yulia and Shuly Wintner. 2010b. Ex- traction of multi-word expressions from small par- allel corpora. In Proceedings of the 23rd Inter- national Conference on Computational Linguistics (COLING 2010), August.", | |
| "links": null | |
| }, | |
| "BIBREF25": { | |
| "ref_id": "b25", | |
| "title": "Semantics-based multiword expression extraction", | |
| "authors": [ | |
| { | |
| "first": "Tim", | |
| "middle": [], | |
| "last": "Van De Cruys", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bego\u00f1a", | |
| "middle": [], | |
| "last": "Villada Moir\u00f3n", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "Proceedings of the Workshop on A Broader Perspective on Multiword Expressions", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "25--32", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Van de Cruys, Tim and Bego\u00f1a Villada Moir\u00f3n. 2007. Semantics-based multiword expression extraction. In Proceedings of the Workshop on A Broader Perspective on Multiword Expressions, pages 25- 32, Prague, Czech Republic, June. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF26": { | |
| "ref_id": "b26", | |
| "title": "Identifying idiomatic expressions using automatic word alignment", | |
| "authors": [ | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Villada Moir\u00f3n", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J\u00f6rg", | |
| "middle": [], | |
| "last": "Bego\u00f1a", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Tiedemann", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2006, | |
| "venue": "Proceedings of the EACL 2006 Workshop on Multi-word-expressions in a multilingual context. Association for Computational Linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Villada Moir\u00f3n, Bego\u00f1a and J\u00f6rg Tiedemann. 2006. Identifying idiomatic expressions using automatic word alignment. In Proceedings of the EACL 2006 Workshop on Multi-word-expressions in a multilin- gual context. Association for Computational Lin- guistics.", | |
| "links": null | |
| }, | |
| "BIBREF27": { | |
| "ref_id": "b27", | |
| "title": "Introduction to the special issue on multiword expressions: Having a crack at a hard nut", | |
| "authors": [ | |
| { | |
| "first": "Aline", | |
| "middle": [], | |
| "last": "Villavicencio", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Francis", | |
| "middle": [], | |
| "last": "Bond", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Anna", | |
| "middle": [], | |
| "last": "Korhonen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Diana", | |
| "middle": [], | |
| "last": "Mccarthy", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Computer Speech & Language", | |
| "volume": "19", | |
| "issue": "4", | |
| "pages": "365--377", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Villavicencio, Aline, Francis Bond, Anna Korhonen, and Diana McCarthy. 2005. Introduction to the special issue on multiword expressions: Having a crack at a hard nut. Computer Speech & Language, 19(4):365-377.", | |
| "links": null | |
| }, | |
| "BIBREF28": { | |
| "ref_id": "b28", | |
| "title": "Definiteness in the Hebrew noun phrase", | |
| "authors": [ | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2000, | |
| "venue": "Journal of Linguistics", | |
| "volume": "36", | |
| "issue": "", | |
| "pages": "319--363", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wintner, Shuly. 2000. Definiteness in the Hebrew noun phrase. Journal of Linguistics, 36:319-363.", | |
| "links": null | |
| }, | |
| "BIBREF29": { | |
| "ref_id": "b29", | |
| "title": "A finite-state morphological grammar of Hebrew", | |
| "authors": [ | |
| { | |
| "first": "Shlomo", | |
| "middle": [], | |
| "last": "Yona", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Shuly", | |
| "middle": [], | |
| "last": "Wintner", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2008, | |
| "venue": "Natural Language Engineering", | |
| "volume": "14", | |
| "issue": "2", | |
| "pages": "173--190", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Yona, Shlomo and Shuly Wintner. 2008. A finite-state morphological grammar of Hebrew. Natural Lan- guage Engineering, 14(2):173-190, April.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "TABREF1": { | |
| "num": null, | |
| "text": "which consists of four sub-corpora: The Knesset corpus contains the Israeli parliament proceedings from 2004-2005; the Haaretz corpus contains articles from the Haaretz newspaper from 1991; The-Marker corpus contains financial articles from the TheMarker newspaper from 2002; and the Arutz 7 corpus contains newswire articles from 2001-2006. Corpora sizes are listed inTable 2.", | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table><tr><td>Corpus</td><td>Number of tokens</td></tr><tr><td>Knesset</td><td>12,742,879</td></tr><tr><td>Harretz</td><td>463,085</td></tr><tr><td>The Marker</td><td>684,801</td></tr><tr><td>Arutz 7</td><td>7,714,309</td></tr><tr><td>Total</td><td>21,605,074</td></tr></table>" | |
| }, | |
| "TABREF2": { | |
| "num": null, | |
| "text": "Corpus data", | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>" | |
| }, | |
| "TABREF3": { | |
| "num": null, | |
| "text": "", | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table><tr><td/><td/><td colspan=\"3\">lists the number of can-</td></tr><tr><td colspan=\"2\">didates in each class.</td><td/><td/><td/></tr><tr><td>Annotator</td><td>+</td><td>-</td><td>0</td><td>err</td></tr><tr><td>1</td><td colspan=\"4\">314 332 238 176</td></tr><tr><td>2</td><td colspan=\"4\">335 403 179 143</td></tr><tr><td>3</td><td colspan=\"3\">400 630 16</td><td>14</td></tr></table>" | |
| }, | |
| "TABREF4": { | |
| "num": null, | |
| "text": "", | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>" | |
| }, | |
| "TABREF7": { | |
| "num": null, | |
| "text": "Optimizing the set of linguisticallymotivated features", | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table/>" | |
| } | |
| } | |
| } | |
| } |