{ "paper_id": "O14-1011", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:04:24.590945Z" }, "title": "Towards automatic enrichment of standardized electronic dictionaries by semantic classes", "authors": [ { "first": "Elleuch", "middle": [], "last": "Imen", "suffix": "", "affiliation": {}, "email": "imen.elleuch@fsegs.rnu.tn" }, { "first": "Gargouri", "middle": [], "last": "Bilel", "suffix": "", "affiliation": {}, "email": "bilel.gargouri@fsegs.rnu.tn" }, { "first": "Ben", "middle": [ "Hamadou" ], "last": "Abdelmajid", "suffix": "", "affiliation": {}, "email": "abdelmajid.benhamadou@isimsf.rnu.tn" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "In this paper we propose an approach for the automatic enrichment of standardized electronic dictionaries by the semantic classes. This approach consists of three phases. The first phase treat the semantic classification process founded on the studies of Gaston Gross. The second phase profites from the existed subject fields in the dictionary's lexical entries in order to attribute the suitable semantic classes. The final phase realizes syntactic analyses of the textual content of meanings's lexical entries. This phase, aims to refine the subject field based enrichment and also treats the non enriched meanings in the second phase. In addition, it attributes the same semantic classes for the synonym meanings. We used an available standardized Arabic dictionary to tested the performance of the proposed approach.", "pdf_parse": { "paper_id": "O14-1011", "_pdf_hash": "", "abstract": [ { "text": "In this paper we propose an approach for the automatic enrichment of standardized electronic dictionaries by the semantic classes. This approach consists of three phases. The first phase treat the semantic classification process founded on the studies of Gaston Gross. The second phase profites from the existed subject fields in the dictionary's lexical entries in order to attribute the suitable semantic classes. The final phase realizes syntactic analyses of the textual content of meanings's lexical entries. This phase, aims to refine the subject field based enrichment and also treats the non enriched meanings in the second phase. In addition, it attributes the same semantic classes for the synonym meanings. We used an available standardized Arabic dictionary to tested the performance of the proposed approach.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "Semantic knowledge, especially semantic classes which aim to characterize meanings of lexical units in dictionaries, have attracted considerable interest in both linguistic (Stede, 1998) , (Dorr, 1997) and computational linguistics (Kipper et al 2000) . Such semantic class can be definite as a semantic linguistic propriety classifying meanings and can therefore be used as a valuable means of comprehending the specific meaning of polysimous lexical units. Thus the need of dictionaries with semantic classes has become a necessity for Natural Language Processing (NLP) applications.", "cite_spans": [ { "start": 173, "end": 186, "text": "(Stede, 1998)", "ref_id": "BIBREF22" }, { "start": 189, "end": 201, "text": "(Dorr, 1997)", "ref_id": "BIBREF3" }, { "start": 232, "end": 251, "text": "(Kipper et al 2000)", "ref_id": "BIBREF16" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "For various languages, various semantic classifications are now available. We can list the verbs classification (Pinker, 1989; Jackendoff, 1990; Levin, 1993, Dubois and Dubois-Charlier, 1997 ) that regroups together verbs that share both a common semantics and a set of syntactic alterna tions. Also, we notice WordNet (Fellbaum, 1998) that provides semantic ontological classification and FrameNet (Fillmore, 1985) that hierarchically classify lexical units using various relationship as synonymy, antonym and is-a relations. However, the referential classification is based on semantic features like [+/-human], [+/concrete], etc. characterizing semantically each lexical unit outside of the meaning's contexts. Object classes (Gross, 1994) defines a semantic classification based on surface realization of predicate argument structure. A semantic class groups together predicates as arguments having the same syntactic constrictions.", "cite_spans": [ { "start": 112, "end": 126, "text": "(Pinker, 1989;", "ref_id": "BIBREF19" }, { "start": 127, "end": 144, "text": "Jackendoff, 1990;", "ref_id": "BIBREF14" }, { "start": 145, "end": 168, "text": "Levin, 1993, Dubois and", "ref_id": null }, { "start": 169, "end": 190, "text": "Dubois-Charlier, 1997", "ref_id": "BIBREF4" }, { "start": 319, "end": 335, "text": "(Fellbaum, 1998)", "ref_id": null }, { "start": 399, "end": 415, "text": "(Fillmore, 1985)", "ref_id": "BIBREF7" }, { "start": 729, "end": 742, "text": "(Gross, 1994)", "ref_id": "BIBREF10" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Rely on a semantic classification; two methods of enrichment lexical resources by semantic classes exist .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "The first one is manual. It is characterized by the large number of lexical units to be classified FrameNet (Fillmore, 1985) this is why it is a costly and time-consuming method. The second method is automatic. It can use corpora (Fuchs & Habert, 2004) , (Condamines, 2005) or in some cases, texts of the treated lexical resources (Rastier, 2001) and (Valette et al, 2006) . The automatic method does not necessitate the intervention of the human expert during the enrichment process (Wilson et al., 2004) . Both manual and automatic method of enrichment lexical resources with semantic classes requires the institution of the semantic classification. In addition, the ability of the structure's dictionary to receive semantic classes is important. In fact, some models of lexical resources do not supply the affectation of the semantic classes to lexical units.", "cite_spans": [ { "start": 108, "end": 124, "text": "(Fillmore, 1985)", "ref_id": "BIBREF7" }, { "start": 230, "end": 252, "text": "(Fuchs & Habert, 2004)", "ref_id": "BIBREF9" }, { "start": 255, "end": 273, "text": "(Condamines, 2005)", "ref_id": "BIBREF1" }, { "start": 331, "end": 346, "text": "(Rastier, 2001)", "ref_id": "BIBREF20" }, { "start": 351, "end": 372, "text": "(Valette et al, 2006)", "ref_id": "BIBREF23" }, { "start": 484, "end": 505, "text": "(Wilson et al., 2004)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "In order to provide a unified framework for modeling lexical resources, in general, and to facilitate the exchange and integration into NLP applications, the LMF (Lexical Markup Framework) standard (Francopoulo & George, 2008) ISO 24613 is published. This standard allows the modelization of all linguistics levels such as the morphological, the syntactic, the semantic and the syntactico-semantic ones.", "cite_spans": [ { "start": 198, "end": 226, "text": "(Francopoulo & George, 2008)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Considering the importance of the semantic classes to characterize the meaning of lexical units, and", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "profiting from the fine model of LMF lexical resources to receive semantic classes, we propose in this paper an automatic approach for the enrichment of standardized LMF electronic dictionaries by semantic classes.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "In fact, the LMF standard offers particular fields (i.e., SubjectField) that can assist the identification of the relevant semantic class and provides synonymy relationships that can be used to improve the enrichment process. Also, in an LMF dictionary, the meaning of lexical entries is accompanied with a rich textual content. The proposed approach is founded on a semantic classification initiated by the Gaston Gross studies. An experimentation of this approach is carried out on an available standardized LMF Arabic dictionary.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "The next part of this paper is organized as follows: We will start with a presentation of some related works related to semantic classification and enrichment methods. Then, we will present the LMF standard.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "Thereafter, we will detail the proposed approach for the enrichment of LMF standardized dictionaries with the semantic classes. After that, we will describe the experiment carried out on a standardized LMF Arabic dictionary and discuss some of the obtained results. Finally, in the conclusion, we will announce some future works.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1." }, { "text": "This section is devoted to the representation of some related works of available semantic classifications and the semantic enrichment methods of lexical resources.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related works", "sec_num": "2." }, { "text": "Several semantic classifications exist in literature. We can mention the verbs classification (Pinker, 1989; Jackendoff, 1990; Levin, 1993 , Dubois & Dubois-Charlier, 1997 . It based on both a common semantics and a set of syntactic alternations to grouped lexical units into semantic classes. This type of classification is restricted to certain class types and treats only verbs. So no comprehensive classification is available limits the usefulness of the class for practical NLP tasks.", "cite_spans": [ { "start": 94, "end": 108, "text": "(Pinker, 1989;", "ref_id": "BIBREF19" }, { "start": 109, "end": 126, "text": "Jackendoff, 1990;", "ref_id": "BIBREF14" }, { "start": 127, "end": 138, "text": "Levin, 1993", "ref_id": "BIBREF17" }, { "start": 139, "end": 171, "text": ", Dubois & Dubois-Charlier, 1997", "ref_id": "BIBREF4" } ], "ref_spans": [], "eq_spans": [], "section": "Semantic classification", "sec_num": "2.1." }, { "text": "Moreover, we can note the ontological classification like WordNet (Miller, 1990 ) that intended to classify philosophical things as they exist in the world. It is particularly appropriate for object modeling, including their relationships and properties. Therefore, content of ontology does not interact directly but rather with relationships (i,e synonymy, antonym, part of, is-A,\u2026). This semantic classification does not consider the use's context of lexical units, further it groups word into classes as presented in the real world without referring to the linguistics features.", "cite_spans": [ { "start": 66, "end": 79, "text": "(Miller, 1990", "ref_id": "BIBREF18" } ], "ref_spans": [], "eq_spans": [], "section": "Semantic classification", "sec_num": "2.1." }, { "text": "Also, we can cite the referential classification (Gross, 1975) (Dichy, 2000) Another kind of semantic classification is proposed by Gaston Gross (Gross, 1994) . It classifies lexical units into semantic classes based on predicate-argument structure. Thus, a semantic class groups together predicates as arguments sharing syntactic and semantic behaviors. Therefore, this classification insures the taking into account the multiple meanings of senses lexical entries depending on a specific use context.", "cite_spans": [ { "start": 49, "end": 62, "text": "(Gross, 1975)", "ref_id": "BIBREF11" }, { "start": 63, "end": 76, "text": "(Dichy, 2000)", "ref_id": "BIBREF2" }, { "start": 145, "end": 158, "text": "(Gross, 1994)", "ref_id": "BIBREF10" } ], "ref_spans": [], "eq_spans": [], "section": "Semantic classification", "sec_num": "2.1." }, { "text": "Finally, we can conclude that the ontological and the referential classification do not guarantee the polysemy of lexical entries because they do not take into account meanings in the classification process. Or the verbs classifications classify only verbs and neglect the other part of speech whereas, the Gaston Gross semantic classification defines a syntactico-semantic classification based on predicate-argument structure. Thus, the variety meaning of senses lexical entries related to an applicable context was ensuring.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Semantic classification", "sec_num": "2.1." }, { "text": "Firstly, the semantic enrichment was done manually. Doing so, this enrichment necessitates high linguist capacities in order to affect the pertinent semantic class to meaning. The LADL tables (Gross, 1975) is one of studies that is based on a manually affectation of semantic features to lexical units meanings.", "cite_spans": [ { "start": 192, "end": 205, "text": "(Gross, 1975)", "ref_id": "BIBREF11" } ], "ref_spans": [], "eq_spans": [], "section": "Semantic enrichment", "sec_num": "2.2." }, { "text": "It is clearly that this manual enrichment is the most relevant one, but it requires a costly time because the vast number of lexical units to be classified and it necessitate the availability of the linguist who attribute the adequate semantic classes to meanings.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Semantic enrichment", "sec_num": "2.2." }, { "text": "With the progress characterizing the computational linguistic domain, the enrichment methods become automatic. This automatic enrichment uses both linguistics features and mathematics techniques to classifying lexical units. This enrichment is marked by three ways. The first uses the linguistic tools for preparing the corpus before classifying lexical units by means of clustering tools (Wilson et al., 2004) . In fact, the construction of the corpus requires the annotation steps that represent a heavy and time consuming task. Several clustering algorithms can be used as Ripper (Cohen, 1996) . The second way uses techniques of automatic clustering (Hatzivassiloglou & McKeown, 1997) . In this case, it is necessary to add syntactic and semantic features in order to achieve the automatic enrichment. The third way consists of using linguistic and statistical approaches. The purpose of this way is to build several types of classifiers and combine their results, either by voting systems or by clustering algorithms (Dziczkowski & Wegrzyn-Wolska, 2008) . This kind of enrichment needs heavier treatments than the other manners listed above.", "cite_spans": [ { "start": 389, "end": 410, "text": "(Wilson et al., 2004)", "ref_id": null }, { "start": 583, "end": 596, "text": "(Cohen, 1996)", "ref_id": "BIBREF0" }, { "start": 654, "end": 688, "text": "(Hatzivassiloglou & McKeown, 1997)", "ref_id": "BIBREF13" }, { "start": 1022, "end": 1058, "text": "(Dziczkowski & Wegrzyn-Wolska, 2008)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Semantic enrichment", "sec_num": "2.2." }, { "text": "LMF is a standard ISO 24613 for modeling lexical knowledge of the majority of natural languages (Francopoulo & George, 2008) . It provides a common model for the representation of electronic lexical resources with guarantees the exchange of data between and among these resources. The LMF model is composed of a core package and a range of extensions referring to the various levels of linguistic analysis (i.e., morphological, syntactic, semantic and syntactico-semantic). The LMF core package describes the basic hierarchy of lexical entry information, including information on the form. The LMF extensions are added to the LMF core components in conjunction with the additional components required for the specific resource modeling. Indeed, to obtain lexical resources according to the LMF standard, it is sufficient to have the core package, then, optionally select packages of extensions necessary to the representation of the modeled dictionary. It is also, essential to select from each extension the corresponding LMF classes required to the treated language. For example, the core package provides the Sense and the Definition classes to describe the meaning of a lexical entry. The MRD (Machine Readable Dictionary) extension reserves the Subject Field class to represent the domain of use of a Sense and the Context class to describe the authentic context for the use of the word form managed by the lexical entry. The LMF semantic extension designates the Sense Relation class to describe the possible relationship between Senses instances such as synonymy and autonomy. Then, the resulting model will be decorated with the Data Categories Registry (DCR) 2 required for the modelization of the dealt language.", "cite_spans": [ { "start": 96, "end": 124, "text": "(Francopoulo & George, 2008)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "LMF standardized model", "sec_num": "3." }, { "text": "In this section, we detail the proposed approach for the automatic enrichment of LMF standardized electronic dictionaries by the semantic classes. The following figure 1 illustrates steps of this approach.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Proposed approach", "sec_num": "4." }, { "text": "The proposed approach is composed of three steps: a semantic classification and two phases of automatic enrichment. To accomplish the aims of the semantic classification, this step requires the hyper-classes of the Gaston Gross classification and a list of verbs and nouns of the studied language in input. The results of the semantic classification step are the ontology of the classification and a list of appropriate verbs and nouns characterizing this classification. Whereas, the SubjectField based enrichment uses the ontology of the classification to enrich the LMF normalized dictionaries by identifying semantic classes. The analysis based enrichment requires achieving the enrichment of the LMF normalized dictionary, both the ontology of the classification and the list of appropriates verbs and nouns identified previously.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 1: Proposed approach", "sec_num": null }, { "text": "Our semantic classification is based on the studies of the Gaston Gross (Gross, 1994) ", "cite_spans": [ { "start": 72, "end": 85, "text": "(Gross, 1994)", "ref_id": "BIBREF10" } ], "ref_spans": [], "eq_spans": [], "section": "Basic concept", "sec_num": "4.1.1." }, { "text": "We propose in figure 2 the general semantic classification process.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Steps of the semantic classification", "sec_num": "4.1.2." }, { "text": "The process of the proposed semantic classification is realized manually by a linguist. It composed by three steps: (i) Adaptation of the classification, (ii) Identification of appropriates verbs and nouns and (iii) Identification of object classes.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 2: Semantic classification", "sec_num": null }, { "text": "Hyper-classes of the Gaston Gross studies (see section 4.1.1) and a list of verbs and nouns of the studied language perform together in order to accomplish the adaptation of the adaptation of the classification step.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Adaptation of the classification:", "sec_num": null }, { "text": "Considering that the semantic classification is performed by a linguist, this step requires the abilities of this expert and the syntactic features of the studied language in order to study the possibility of the adaptation of the semantic classification on the studied language. On the basis of syntactic features of the studied language, the expert can identify new hyper-classes appropriate to the treated language, delete or rename the existing semantic hyper-classes. Therefore the compliant hyper-classes represent the result of this step.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Adaptation of the classification:", "sec_num": null }, { "text": "ii.Identification of appropriates verbs and nouns:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Adaptation of the classification:", "sec_num": null }, { "text": "On the basis of the novel list of hyper-classes identified in the previous step, related to the specific studied language, the identification of appropriates verbs and nouns take place. This step aims to detect the appropriate list of verbs and nouns characterizing each hyper-classes of the proposed semantic classification.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Adaptation of the classification:", "sec_num": null }, { "text": "The object class concept represents the characteristic of the proposed semantic classification. Thus, the aim of this step is the identification of object classes for each semantic class. To accomplish this objective, this step requires the compliant hyper-classes of the studied language and the list of appropriates verbs and nouns recognized in the last step. The results of this step affect predicates-semantic classes to as well as arguments.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "iii. Identification of object classes:", "sec_num": null }, { "text": "Indeed, the expert benefits from the syntactic features of the studied language in order to identify objects classes relating respectively to hyper-semantic classes of predicates and arguments. As hyper-classes, the identification of the object classes outcomes a list of verbs and nouns characterizing each object class. This list performs to update the list of appropriates verbs and nouns of the classification. An ontology of the classification that's regroups all complaint hyper-classes and object classes related to the studied language represent the result of this step.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "iii. Identification of object classes:", "sec_num": null }, { "text": "After developing a semantic classification, the enrichment process of the standardized LMF dictionaries with semantic classes will take place. It composed of two main phases: (i) the Subject Field based enrichment that benefited from the LMF dictionaries structure, particularly from the uses domains related to meanings of lexical entries (ii) the analysis based enrichment that uses features of the obtained semantic classification.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Enrichment of LMF standardized dictionaries", "sec_num": "4.2." }, { "text": "This enrichment is based on the field \"SubjectField\" according to the LMF model. As shown in figure 3, it consists of two steps described as follow: i. Searching senses with \"SubjectField\": the domains of uses for each \"Senses\" of lexical entries in LMF normalized dictionary are represented through a class named \"SubjectField\". The aim of this step is the extraction from the dictionary, Senses related to treated lexical entry containing the \"SubjectField\" field.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Subject Field based enrichment", "sec_num": "4.2.1." }, { "text": "ii. Identification of semantic classes: a pretreatment realized on the obtained semantic classification and the existed \"Subjectfield\" in an LMF stadardized dictionary can made a directly correspondence between the hyper-semantic classes or the object classes with the \"SubjectField\". If this is the case, this step identifies the semantic class from the ontology of the classification related to the founded \"SubjectField\" and updates the LMF standardized dictionary by the addition of the retained semantic class to the corresponding Sense. i. Searching enriched senses: this step aims to search from LMF normalized dictionary the enriched senses with semantic classes based on the SubjectField based enrichment and in the same time the nonenriched senses. A specific treatment will be affected to those senses in the next step.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Subject Field based enrichment", "sec_num": "4.2.1." }, { "text": "ii. Generation of restricted appropriates verbs and nouns: the assignment of the semantic class identified by the SubjectField based enrichment is not a definitive assignment. Indeed, in order to achieve the definitive enrichment, this step requires for the realization of its process both the Appropriates verbs and nouns and the which will be affected to the Sense in order to enrich semantically the LMF dictionary. v. Synonymy based enrichment: in this step we have identified and affected a semantic class to the treated Sense. After that, the synonymy based enrichment takes place, it aims to search the synonymy senses related to the treated sense. Then, the same semantic class identified by the exhaustive or the refined enrichment will be affected to the synonymy senses. At the end of this step, we obtain an enriched sense with the relevant semantic class and also the related synonymy senses enriched by the same semantic class.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Subject Field based enrichment", "sec_num": "4.2.1." }, { "text": "This section focuses an experimentation of the proposed approach of the automatic enrichment of standardized dictionaries by semantic classes. An Arabic LMF dictionary is used to test the performance of this approach.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation on the Arabic language", "sec_num": "5." }, { "text": "With respect to the Arabic language and to our knowledge there has been no works treated effectively an Arabic semantic classification. In fact, in literature available works are limited to some attempts of specialized dictionaries without related to any theoretical semantic classification. We can note for example, the \" \u202b\u0627\u0644\u0644\u063a\u062d\u202c \u202b\u0641\u0642\u0648\u202c \u202b\u0627\u0644\u0639\u0632\u062a\u064a\u062d\u202c \u202b/\u064b\u0633\u0632\u202c fiq.hu all~u\u03b3a\u0127i wa sir~u al\u03c2arabiy~ati\" dictionary created by \" \u202b\u0627\u0644\u062b\u0639\u0627\u0644\u062b\u064a\u202c \u202b\u0647\u0646\u0635\u064c\u0631\u202c \u202b/\u0623\u062a\u064c\u202c Aabuw mansuwr al\u03b8~a\u03c2aAlibiy\" which groups lexical units into thirty chapters. Each chapter is subdivided into sub-chapters grouping together lexical units sharing the same semantic meaning. The chapter \" \u202b\u064b\u0647\u0627\u202c \u202b\u064b\u0627\u0644\u0633\u0627\u0644\u062d\u202c \u202b\u062a\u0648\u202c \u202b\u064a\u062a\u0635\u0644\u202c \u202b\u064b\u0647\u0627\u202c \u202b\u0627\u0644\u0644\u062b\u0627\u0633\u202c \u202b\u0641\u064a\u202c \u202b\u0627\u0623\u0644\u062f\u064b\u0627\u062e\u202c \u202b\u064b\u0633\u0627\u0626\u0632\u202c \u202b\u0625\u0644\u064a\u0648\u202c \u202b\u064a\u0646\u0636\u0627\u0641\u202c \u202b\u0647\u0623\u062e\u0630\u0649\u0627\u202c \u202b\u064a\u0623\u062e\u0630\u202c \u202b\u064b\u0647\u0627\u202c \u202b\u064b\u0627\u0622\u0644\u0627\u0644\u062e\u202c / fiy al~ibaAs wa maA yat~asilu bihi wa als~ilaAH wa maA yan.DaAfu \u01cdilay.hi wa saA\u0177iri alAadawaAti wa al\u0100laAti wa maA yuA.xaDu maA.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Choice of the Arabic language", "sec_num": "5.1." }, { "text": "xaDahaA\" includes fortynine sub-chapters as \u202b\u0627\u0644\u0646\u0633\u064a\u062c\"\u202c \u202b\u062a\u0642\u0633\u064a\u0646\u202c \u202b\u0641\u064a\u202c /fiy taq.siym aln~asiyj\" \"the division of tissues\", \u202b\u0627\u0644\u062e\u064a\u0627\u0637\u062d\"\u202c \u202b\u062a\u0642\u0633\u064a\u0646\u202c \u202b/\u0641\u064a\u202cfiy taq.siym alHiyaATa\u0127i \" \"the division of sewing \", \" \u202b\u064b\u062a\u0641\u0635\u064a\u0644\u064a\u0627\u202c \u202b\u0627\u0644\u062e\u064a\u064c\u0637\u202c \u202b\u062a\u0642\u0633\u064a\u0646\u202c \u202b/\u0641\u064a\u202c fiy taq.siym alxuyuwT wa tafSiyluhaA \" \"the division of thread and its peculiarities \"....", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Choice of the Arabic language", "sec_num": "5.1." }, { "text": "\" \u202b\u0627\u0644\u0648\u0627\u0644\u062a\u0633\u202c \u202b\u0627\u0644\u0633\u0648\u0627\u0621\u202c \u202b\u0627\u0644\u0639\u0632\u062a\u064a\u202c \u202b/\u0627\u0644\u0648\u0639\u062c\u0646\u202calmu\u03c2.jam al\u03c2arabiy lias.maA'i almalaAbis\" is another Arabic dictionary specialized in the classification of Arabic nouns of clothes. \" \u202b\u0627\u062a\u0632\u0627\u0649\u064a\u0646\u202c \u202b\u0639\u062b\u0630\u202c \u202b/\u0631\u062c\u0629\u202c rajab \u03c2abd Aib.raAhiym\" the writer of this lexical resource grouped more than 1250 clothes Arabic nouns.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Choice of the Arabic language", "sec_num": "5.1." }, { "text": "In this section, we experiment the process of the semantic classification (see section 4.1) on the Arabic language. We were interested in this experimentation on the \"CONCRETE\" hyper-class of arguments. This hyper-class is also retained for the Arabic language from the classification of Gaston Gross. Among the object classes belonging to the \"CONCRETE\" hyper class we note the \"Clothes\" class. Indeed, the Arabic verb \u202b\"\u0644\u062b\u0633\"\u202c \"to wear\", represent the appropriate verb characterizing this object class. Thus, one meaning of this verb describe an \"ACTION\" realized by a first \"HUMAN\" argument and highlighting another \"CONCRETE\" argument. The example bellow illustrates three sentences detailed the mean of the \u202b\"\u0644\u062b\u0633\"\u202c \"to wear\" verb:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "( Indeed, in sentence (1), a \"pupil\" can \"wear a hat\", while in sentence (2) a \"Pupil\" cannot wear an \"apple\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "because an \"apple\" is an \"Aliment\" so it can be eaten but not wean. Whereas, in sentence (3) the \"water\" is an \"Aliment/ water\" and cannot be wean. Those examples explicate the requirement of the creation of the \"Clothes\" and the \"Aliment\" objects classes under the \"CONCRETE\" hyper-class. Thus, he \"Clothes\" object class includes all nouns that can be worn by a \"HUMAN\". Arabic verbs such as: \u202b\u062e\u0644\u0639\"\u202c /xala\u03c2a/to undress\" \"\u064d\u202b/\u0627\u0631\u062a\u0630\u202cAir.tady'/to dress\",\" \u202b\u0644\u062b\u0633\u202c /labisa/to wear\", and nouns like: \"\u202b/\u0645\u0633\u0627\u0621\u202ckisaA''/cloth\", \"\u202b/\u0644\u062b\u0627\u0633\u202clibaAs/wear\", \"\u202b/\u062b\u064c\u0628\u202c\u03b8aw.jb\u0169/dress\" characterize the \"Clothes\" object class.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "Arguments instances of the \"Clothes\" object class can be: \"\u202b/\u0627\u0630\u0627\u0621\u202cHidaA'/shoes\", \"\u202b/\u0646\u0639\u0644\u202cna\u03c2.l/sock\", \"\u202b/\u062e\u0641\u202cxuf~\u0169/slipper\", \u202b\u062b\u0639\u062d\"\u202c /qub~a\u03c2a\u0127\u0169/hat\", \"\u202b/\u0633\u0632\u064b\u0627\u0644\u202csir.waAl\u0169/pant\", \u202b\u0648\u064a\u0635\"\u202c /qamiuS\u0169/shirt\". Thus, the appropriate verbs of the \"Clothes\" object class like \u202b\u062e\u0644\u0639\"\u202c /xala\u03c2a/to undress\" \"\u064d\u202b/\u0627\u0631\u062a\u0630\u202cAir.tady'/to dress\"\" \u202b/\u0644\u062b\u0633\u202clabisa/to wear\" can be correctly introduces the arguments instances list before. But in Arabic language, some verbs select from the \"Clothes\" arguments instances a specific ones but cannot use all of them. For example, the \u202b/\u0627\u0627\u062a\u0630\u064d(\u202c Ain.ta\u03c2ala / to wear shoes, \u202b\u0627\u0646\u062a\u0639\u0644\u202c /AiH.tady /to wear shoes) verbs cannot precede allof the arguments instances \"Clothes\" but only \u202b.>\u0627\u0644\u062d\u0630\u0627\u0621<\u202c Thus, the sentence ( \u202b\u0627\u0644\u0642\u0648\u064a\u0635\u202c \u202b)\u0627\u0627\u062a\u0630\u064d\u202c (he wear shoes shirt) is semantically incorrect because (\u064d\u202b/\u0627\u0627\u062a\u0630\u202cAin.ta\u03c2ala / to wear shoes) is an appropriate verb to \u202b>\u0627\u0644\u062d\u0630\u0627\u0621<\u202c class and \u202b/\u0627\u0644\u0642\u0648\u064a\u0635(\u202c alqamiuS\u0169/shirt) does not represent \u202b>\u0627\u0644\u062d\u0630\u0627\u0621<\u202c but rather \u202b\u062b\u064a\u0627\u0628<\u202c >.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "Therefore, it is necessary to create two objects classes under the \"Clothes\" namely \u202b>\u0627\u0644\u062d\u0630\u0627\u0621<\u202c and < fringues > \u202b.>\u062b\u064a\u0627\u0628<\u202c", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "The table 1 in following summarizes the previous idea:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "Table1: Appropriate verbs for the \u202b>\u0627\u0644\u062d\u0630\u0627\u0621<\u202c and \u202b>\u062b\u064a\u0627\u0628<\u202c object class", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Classification of Arabic arguments", "sec_num": "5.2.1." }, { "text": "In this section we present an example of the semantic classification ontology for Arabic language. The below figure 5 illustrates some recognized hyper-classes and object classes using the proposed process of the semantic classification (see section 4.1). This figure is created with the OWL ontology.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Results of the Arabic semantic classification", "sec_num": "5.2.2." }, { "text": "As presented in the figure 5, a \"CONCRETE\" is a hyper class. Among the object classes founded under the \"CONCRETE\" hyper class we note the \"Clothes\" and the \"Aliments\" subclasses. The \"Clothes\" object class is subdivided into the following object-sub-classes: \"Head_wear\", \"Fringues\", \"Shoes\" and \"Others_Clothes\".", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure5: Examples of Arabic hyper-classes and object classes of arguments", "sec_num": null }, { "text": "And so on for the other hyper and object classes. Table 2 present the list of appropriate verbs and nouns related to the \"CONCRETE\" class:", "cite_spans": [], "ref_spans": [ { "start": 50, "end": 57, "text": "Table 2", "ref_id": null } ], "eq_spans": [], "section": "Figure5: Examples of Arabic hyper-classes and object classes of arguments", "sec_num": null }, { "text": "Table2: Appropriate verbs and nouns of the object class", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure5: Examples of Arabic hyper-classes and object classes of arguments", "sec_num": null }, { "text": "The Arabic LMF standardized dictionary is a lexical resource conforms to the LMF standard ISO-24613. The model of this dictionary (khemakhem et al 2013) covers all lexical levels: morphological, syntactic, semantic and syntactico-semantic. This dictionary contains about 37000 lexical entries among them 10800 verbs and 3800 roots.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Arabic LMF standardized dictionary", "sec_num": "5.3.1." }, { "text": "In Arabic LMF standardized dictionary, three classes namely Definition, Context The application of the process of the refined enrichment by using the restricted list of verbs and nouns (table5 in yellow) on the last extracted fragment used in the SubjectField based enrichment (figure 6) can give the enrichment presented in the following figure 7.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation of the \"SubjectField\" based enrichment", "sec_num": "5.3.2." }, { "text": "In the following, we present an experimentation of the exhaustive enrichment using the appropriate verbs and nouns applied to non-enriched senses. The analysis of Contexts and Definitions of senses related to a lexical entry in the Arabic LMF standardized dictionary by using the appropriate verbs and nouns (table4) can identify the relevant semantic class. ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 7: The refined enrichment applied to a sense of lexical entry", "sec_num": null }, { "text": "To test the performance of the carried out experimentation, we have realized a statistical evaluation.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Results", "sec_num": "5.4." }, { "text": "Our Arabic standardized dictionary contains in total 34000 lexical entries including 62157 senses. Concerning the SubjectField based enrichment experimentation; we have used 4 \"SubjectField\" (Animal, Insect, Plant and Culinary) among the 19 available in the Arabic dictionary. And for the analysis based enrichment we have choice only the \"CONCRETE\" hyper class and specially the \"Clothes\" object class to experiment the process of this kind of enrichment.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Results", "sec_num": "5.4." }, { "text": "The table 5 below gives the statistical evaluation of the \"SubjectField\" and the analysis based enrichment. ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Results", "sec_num": "5.4." }, { "text": "In this paper, we have proposed an approach for the automatic enrichment of LMF standardized dictionaries with semantic classes. This approach is composed of a semantic classification based on the Gaston Gross studies and two types of enrichment. The first enrichment named SubjectField based enrichment, takes advantages from the structure of an LMF dictionary where meanings contain the domain of use of a lexical entry. The second enrichment called analysis based enrichment, uses the features of the proposed semantic classification based on appropriates verbs and nouns specifying each semantic class and applied to the available text components in the dictionary.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion and perspectives", "sec_num": "6." }, { "text": "We carried out experimentation, by using an available Arabic standardized dictionary. The obtained results are satisfying concerning the SubjectField based enrichment. The synonymy based enrichment can reduce the enrichment effort at thirds because on average, the synonymy relation connects three or more senses.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion and perspectives", "sec_num": "6." }, { "text": "In the future, we opted to achieve the experimentation on the others semantic classes of the proposed semantic classification for Arabic language and to complete the rest of SubjectField existed in the Arabic LMF standardized dictionary. In addition, we consider improving the analysis based enrichment by adding more efficient syntactic-semantic analysis. Finally, we foresee that the enrichment can offer the flexibility to create new oriented versions of the semantic knowledge needed for different NPL applications.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusion and perspectives", "sec_num": "6." }, { "text": "www.isocat.org", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [ { "text": "SubjectIn the Arabic LMF normalized dictionary, the \u202b\"\u0627\u064a\u064c\u0627\u0649\"\u202c \"Animal\" \"HayawaAn\" and the \u202b\"\u0627\u0634\u0632\u062c\"\u202c \"Ha\u0161ara\u0127\"\"Insect\" SubjectFields can be grouped into the \"animal\" hyper-class. It is important to indicate that the \u202b\"\u0627\u0634\u0632\u062c\"\u202c \"Ha\u0161ara\u0127\" \"Insect\" SubjectField corresponds directly to the object class named \"Insect\" and the \u202b\"\u0627\u064a\u064c\u0627\u0649\"\u202c \"HayawaAn\" \"animal\" SubjectField may correspond to the object classes: \"Bird\", \"Rodents\", \"reptiles\" and \"Aquatic-animals\" as shows in figure5.The following figure 6 illustrates an example of the SubjectField based enrichment.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Table3: Examples of available Subject Field in the Arabic standardized dictionary", "sec_num": null }, { "text": "The analysis based enrichment is subdivided into two kind of enrichment. The first enrichment appointed refined enrichment requires for the progress of its process the restricted list of verbs and nouns in order to refine the primary enrichment carried out in the SubjectField based enrichment. Or the second enrichment is exhaustive, concerning only non-enriched senses, uses the appropriate verbs and nouns of the semantic classification in order to realize the semantic enrichment of the dictionary.The table 4 given in the following, contains the restricted list of appropriates verbs and nouns related to the \u202b\"\u0627\u064a\u064c\u0627\u0649\"\u202c \"HayawaAn\" \"animal\" arguments hyper-class.The \u202b\"\u0627\u064a\u064c\u0627\u0649\"\u202c \"HayawaAn\" \"animal\" SubjectField Added Semantic class", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation of the analysis based enrichment", "sec_num": "5.3.3." } ], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Law, social policy, and violence: The impact of regional cultures", "authors": [ { "first": "D", "middle": [], "last": "Cohen", "suffix": "" } ], "year": 1996, "venue": "Journal of personality and Social Psychology", "volume": "70", "issue": "", "pages": "961--978", "other_ids": {}, "num": null, "urls": [], "raw_text": "Cohen, D. (1996). Law, social policy, and violence: The impact of regional cultures. Journal of personality and Social Psychology, 70.961-978.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "S\u00e9mantique et corpus, quelles rencontres possibles ? S\u00e9mantique et corpus, A. Condamines, \u00e9d", "authors": [ { "first": "A", "middle": [], "last": "Condamines", "suffix": "" } ], "year": 2005, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Condamines A. (2005). S\u00e9mantique et corpus, quelles rencontres possibles ? S\u00e9mantique et corpus, A. Condamines, \u00e9d., Paris : Herm\u00e8s.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Morphosyntactic Specifieers to be associated to Arabic lexical Entries-Methodological and Theoretical Aspects", "authors": [ { "first": "J", "middle": [], "last": "Dichy", "suffix": "" } ], "year": 2000, "venue": "Corpora ans Natural Language Processing", "volume": "", "issue": "", "pages": "55--60", "other_ids": {}, "num": null, "urls": [], "raw_text": "Dichy, J. (2000). Morphosyntactic Specifieers to be associated to Arabic lexical Entries-Methodological and Theoretical Aspects. Proceddings of ACIDA'2000. Monastir, Tunisia, 22-24 March 2000. Corpora ans Natural Language Processing. Volume, p.55-60.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Large-scale dictionary construction for foreign language tutoring and interlingual machine translation", "authors": [ { "first": "B", "middle": [], "last": "Dorr", "suffix": "" } ], "year": 1997, "venue": "Machine Translation", "volume": "12", "issue": "4", "pages": "271--325", "other_ids": {}, "num": null, "urls": [], "raw_text": "Dorr, B. (1997). Large-scale dictionary construction for foreign language tutoring and interlingual machine translation. Machine Translation, 12(4):271-325.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Les verbes fran\u00e7ais (LVF) . Jean Dubois et Dubois-Charlier Fran\u00e7oise Diffuseur exclusif. Larousse Bordas", "authors": [ { "first": "J", "middle": [], "last": "Dubois", "suffix": "" }, { "first": "F", "middle": [], "last": "Dubois-Charlier", "suffix": "" } ], "year": 1997, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Dubois, J. & Dubois-Charlier, F. (1997). Les verbes fran\u00e7ais (LVF) . Jean Dubois et Dubois-Charlier Fran\u00e7oise Diffuseur exclusif. Larousse Bordas, Paris.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "An autonomous system designed for automatic detection and rating of film reviews. Extraction and linguistic analysis of sentiments", "authors": [ { "first": "G", "middle": [], "last": "Dziczkowski", "suffix": "" }, { "first": "K", "middle": [], "last": "Wegrzyn-Wolska", "suffix": "" } ], "year": null, "venue": "IEEE/WIC/ACM International Conferences on Web Intelligence. WI'08", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Dziczkowski, G. & Wegrzyn-Wolska, K. An autonomous system designed for automatic detection and rating of film reviews. Extraction and linguistic analysis of sentiments. In IEEE/WIC/ACM International Conferences on Web Intelligence. WI'08, Australie.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "The organization of verbs and verb concepts in a semantic net", "authors": [ { "first": "C", "middle": [], "last": "Fellbaum", "suffix": "" } ], "year": 1999, "venue": "Predicative Forms in Natural Language and in Lexical Knowledge Bases", "volume": "", "issue": "", "pages": "93--110", "other_ids": {}, "num": null, "urls": [], "raw_text": "Fellbaum, C. (1999). The organization of verbs and verb concepts in a semantic net. In P. Saint-Dizier, editor, Predicative Forms in Natural Language and in Lexical Knowledge Bases, pages 93-110. Kluwer Academic Publishers, Netherlands.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Frame and the semantics of understanding.Quaderni di semantica", "authors": [ { "first": "C", "middle": [ "J" ], "last": "Fillmore", "suffix": "" } ], "year": 1985, "venue": "", "volume": "6", "issue": "", "pages": "222--254", "other_ids": {}, "num": null, "urls": [], "raw_text": "Fillmore, C.J. (1985). Frame and the semantics of understanding.Quaderni di semantica, 6(2):222-254.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "ISO/TC 37/SC 4 Rev.16. Language resource management-Lexical markup framework (LMF)", "authors": [ { "first": "G", "middle": [], "last": "Francopoulo", "suffix": "" }, { "first": "M", "middle": [], "last": "George", "suffix": "" } ], "year": 2008, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Francopoulo, G. & George, M. (2008). ISO/TC 37/SC 4 Rev.16. Language resource management-Lexical markup framework (LMF).", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "Traitement automatique et ressources num\u00e9ris\u00e9es pour le fran\u00e7ais, Le fran\u00e7ais moderne", "authors": [ { "first": "C", "middle": [], "last": "Fuchs", "suffix": "" }, { "first": "B", "middle": [], "last": "Habert", "suffix": "" } ], "year": 2004, "venue": "", "volume": "72", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Fuchs, C., Habert, B., \u00e9d. (2004). Traitement automatique et ressources num\u00e9ris\u00e9es pour le fran\u00e7ais, Le fran\u00e7ais moderne, Vol. 72, n\u00b01.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Classes d'objets et description des verbes", "authors": [ { "first": "G", "middle": [], "last": "Gross", "suffix": "" } ], "year": 1994, "venue": "Langages", "volume": "115", "issue": "", "pages": "15--30", "other_ids": {}, "num": null, "urls": [], "raw_text": "Gross, G. (1994). Classes d'objets et description des verbes. Langages, 115, 15-30.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "M\u00e9thodes en syntaxe : R\u00e9gimes des constructions compl\u00e9tives", "authors": [ { "first": "M", "middle": [], "last": "Gross", "suffix": "" } ], "year": 1975, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Gross, M. (1975). M\u00e9thodes en syntaxe : R\u00e9gimes des constructions compl\u00e9tives. Hermann, Paris, France.", "links": null }, "BIBREF12": { "ref_id": "b12", "title": "Arabic Computational Morphology: Knowledgebased and Empirical Methods", "authors": [ { "first": "N", "middle": [], "last": "Habash", "suffix": "" }, { "first": "A", "middle": [], "last": "Soudi", "suffix": "" }, { "first": "T", "middle": [], "last": "Buckwalter", "suffix": "" } ], "year": 2007, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Habash, N., Soudi, A., Buckwalter, T., et al. (2007). In Arabic Computational Morphology: Knowledge- based and Empirical Methods. ISBN: 978-1-4020-6045-8", "links": null }, "BIBREF13": { "ref_id": "b13", "title": "Predicting the Semantic Orientation of Adjectives", "authors": [ { "first": "V", "middle": [], "last": "Hatzivassiloglou", "suffix": "" }, { "first": "K", "middle": [], "last": "Mckeown", "suffix": "" } ], "year": 1997, "venue": "ACL '98 Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and Eighth Conference of the European Chapter of the Association for Computational Linguistics", "volume": "", "issue": "", "pages": "174--181", "other_ids": {}, "num": null, "urls": [], "raw_text": "Hatzivassiloglou, V., McKeown, K. (1997). Predicting the Semantic Orientation of Adjectives. ACL '98 Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and Eighth Conference of the European Chapter of the Association for Computational Linguistics. Pages 174-181.", "links": null }, "BIBREF14": { "ref_id": "b14", "title": "Semantic Structures", "authors": [ { "first": "R", "middle": [], "last": "Jackendoff", "suffix": "" } ], "year": 1990, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Jackendoff, R. (1990). Semantic Structures. MIT Press, Cambridge, Massachusetts.", "links": null }, "BIBREF15": { "ref_id": "b15", "title": "LMF for Arabic, chapter in the book \"LMF:Lexical Markup Framework", "authors": [ { "first": "A", "middle": [], "last": "Khemakhem", "suffix": "" }, { "first": "B", "middle": [], "last": "Gargouri", "suffix": "" }, { "first": "K", "middle": [], "last": "Haddar", "suffix": "" }, { "first": "A", "middle": [], "last": "Ben Hamadou", "suffix": "" } ], "year": 2013, "venue": "", "volume": "", "issue": "", "pages": "83--96", "other_ids": {}, "num": null, "urls": [], "raw_text": "Khemakhem, A., Gargouri, B., Haddar K. Ben Hamadou, A, et al. (2013). LMF for Arabic, chapter in the book \"LMF:Lexical Markup Framework\". Wiley Editions, ISBN: 9781848214309, pp.83-96, March 2013.", "links": null }, "BIBREF16": { "ref_id": "b16", "title": "Classbased construction of a verb lexicon", "authors": [ { "first": "K", "middle": [], "last": "Kipper", "suffix": "" }, { "first": "H", "middle": [ "T" ], "last": "Dang", "suffix": "" }, { "first": "M", "middle": [], "last": "Palmer", "suffix": "" } ], "year": 2000, "venue": "Proc. of the 17th National Conference on Artificial Intelligence", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kipper, K., Dang, H. T., Palmer, M., et al. (2000). Classbased construction of a verb lexicon. In Proc. of the 17th National Conference on Artificial Intelligence. Austin, TX.", "links": null }, "BIBREF17": { "ref_id": "b17", "title": "English Verb Classes and Alternations", "authors": [ { "first": "B", "middle": [], "last": "Levin", "suffix": "" } ], "year": 1993, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Levin, B. (1993). English Verb Classes and Alternations. Chicago University Press, Chicago.", "links": null }, "BIBREF18": { "ref_id": "b18", "title": "WordNet: An on-line lexical database", "authors": [ { "first": "G", "middle": [ "A" ], "last": "Miller", "suffix": "" } ], "year": 1990, "venue": "International Journal of Lexicography", "volume": "3", "issue": "4", "pages": "235--312", "other_ids": {}, "num": null, "urls": [], "raw_text": "Miller, G. A. (1990). WordNet: An on-line lexical database. International Journal of Lexicography, 3(4):235-312.", "links": null }, "BIBREF19": { "ref_id": "b19", "title": "Learnability and Cognition: The Acquisition of Argument Structure", "authors": [ { "first": "S", "middle": [], "last": "Pinker", "suffix": "" } ], "year": 1989, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Pinker, S. (1989). Learnability and Cognition: The Acquisition of Argument Structure. MIT Press, Cambridge, Massachusetts.", "links": null }, "BIBREF20": { "ref_id": "b20", "title": "Arts et sciences du texte", "authors": [ { "first": "F", "middle": [], "last": "Rastier", "suffix": "" } ], "year": 2001, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Rastier, F. (2001). Arts et sciences du texte. Paris : PUF.", "links": null }, "BIBREF21": { "ref_id": "b21", "title": "BoosTexter: A Boosting-based System for Text Categorization", "authors": [ { "first": "R", "middle": [ "E" ], "last": "Schapire", "suffix": "" }, { "first": "Y", "middle": [], "last": "Singer", "suffix": "" } ], "year": 2000, "venue": "Machine Learning", "volume": "39", "issue": "", "pages": "135--168", "other_ids": {}, "num": null, "urls": [], "raw_text": "Schapire, R. E. & Singer, Y. (2000). BoosTexter: A Boosting-based System for Text Categorization. Machine Learning, 39, 135-168, 2000", "links": null }, "BIBREF22": { "ref_id": "b22", "title": "A generative perspective on verb alterntions", "authors": [ { "first": "M", "middle": [], "last": "Stede", "suffix": "" } ], "year": 1998, "venue": "Computational Linguistics", "volume": "24", "issue": "3", "pages": "401--430", "other_ids": {}, "num": null, "urls": [], "raw_text": "Stede, M. (1998). A generative perspective on verb alterntions. Computational Linguistics, 24(3):401-430.", "links": null }, "BIBREF23": { "ref_id": "b23", "title": "Pour une approche s\u00e9mique du sens. Verbum ex machina, Actes de la 13\u00e8me conf\u00e9rence sur le traitement automatique des langues naturelles", "authors": [ { "first": "M", "middle": [], "last": "Valette", "suffix": "" }, { "first": "A", "middle": [], "last": "Estacio-Moreno", "suffix": "" }, { "first": "E", "middle": [], "last": "Petitjean", "suffix": "" }, { "first": "E", "middle": [], "last": "Jacquey", "suffix": "" } ], "year": 2006, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Valette, M., Estacio-Moreno, A., PetitJean, E., Jacquey, E. (2006). El\u00e9ments pour la g\u00e9n\u00e9ration de classes s\u00e9mantiques \u00e0 partir de d\u00e9finitions lexicographiques. Pour une approche s\u00e9mique du sens. Verbum ex machina, Actes de la 13\u00e8me conf\u00e9rence sur le traitement automatique des langues naturelles (TALN 06).", "links": null }, "BIBREF25": { "ref_id": "b25", "title": "Twelve Ways to Make Sure Your Parallel Programming System Doesn't Make Others Look Bad", "authors": [ { "first": "G", "middle": [ "V" ], "last": "Wilson", "suffix": "" }, { "first": "B", "middle": [], "last": "Gorda", "suffix": "" }, { "first": "P", "middle": [], "last": "Lu", "suffix": "" } ], "year": 1994, "venue": "IEEE Computer", "volume": "27", "issue": "10", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Wilson, G.V., Gorda, B., Lu, P., et al. (1994). Twelve Ways to Make Sure Your Parallel Programming System Doesn't Make Others Look Bad. IEEE Computer, 27(10), 1994.", "links": null } }, "ref_entries": { "FIGREF0": { "uris": null, "text": "that used semantic features like [+/concrete], [+/-human]. Those features are attached to lexical units to describe their appurtenance to the semantic classes. This semantic classification assigned semantic features to lexical entries without taking into account the uses of the lexical units.", "type_str": "figure", "num": null }, "FIGREF1": { "uris": null, "text": "semantic classification (see section 2). This classification uses the predicate-argument structure to classify lexical units. Thus, the simple sentence represents the minimum unit of analysis. Indeed, two major semantic classes characterize this classification namely: the semantic classes of predicates and the semantic classes of arguments. However, prior to the object classes, and based on syntactic features, the classification maintains classes that regroup all predicates that share the common syntactic behaviors named Hyper-classes. Thus, hyper-classes of predicates, specified by this classification are: \"ACTION, EVENT, STATE and PREDICATIVE HUMAN.\" While hyperclasses of arguments are: \"HUMAN, CONCRETE, PLANTS, ANIMALS, TIME, RENTAL and ABSTRACT.\" These hyper-classes are subject to sub classifications by means of arguments permutations (distributional criteria) appearing in one or more positions of arguments related to a given predicate. Thus, if a permutation of a noun by another contributes to a rupture of the meaning of a predicate sense, then a new object class is required to be created. These object classes allow highlighting the different uses of a polysemous predicate.", "type_str": "figure", "num": null }, "FIGREF2": { "uris": null, "text": "Subject Field based enrichment 4.2.2. The analysis based enrichment The analysis based enrichment uses the features of the retained semantic classification. The following figure 4 illustrates the steps of this kind of enrichment.", "type_str": "figure", "num": null }, "FIGREF3": { "uris": null, "text": "Analysis based enrichment", "type_str": "figure", "num": null }, "FIGREF4": { "uris": null, "text": "Enriched senses based on SubjectField. The restricted appropriates verbs and nouns represent the result of the generation of restricted appropriates verbs and nouns phase.iii. Refined enrichment: the restricted list of verbs and nouns identified in the last step, the sense alreadyenriched based on SubjectField and the ontology of the classification represents the input of this step. Indeed, this step uses the restricted list of verbs and nouns to analyze the textual content of the enriched \"Sense\" in order to refine the semantic class assignment. Thus a relevant semantic class is identified based on the ontology of the classification and will definitive be attribute to the treated Sense. iv. Exhaustive enrichment: the exhaustive enrichment concerns the non-enriched senses. In fact, a specific treatment is performed to those non-enriched senses by the means of the Appropriates verbs and nouns identified by the retained semantic classification. This treatment consist of an analyze of the \"Contexts\" and the \"Definitions\" field related to a Sense of a lexical entry in the LMF dictionary using the appropriate verbs and nouns. This analyze identify the relevant semantic class from the ontology of the semantic classification", "type_str": "figure", "num": null }, "FIGREF5": { "uris": null, "text": "The exhaustive enrichment applied to a sense of lexical entry Enriched sense based on \"SubjectField\" Appropriate noun of the object class Enriched sense based on refined enrichment An appropriate verbs of the Object class An appropriate noun of the < Other wear > Sub-Object-class Semantic class", "type_str": "figure", "num": null }, "TABREF1": { "type_str": "table", "content": "
Table4: Restricted list of appropriate verbs and nouns of \"Animal\" hyper-class
Hyper classRestricted list of appropriate nouns and verbsObject classesRestricted list of verbs appropriate nouns andSub-Object-classes
\u202b\u0627\u0648\u0627\u0645\u202c\u202b\u0639\u0635\u0641\u064c\u0631\u202c
\u202b\u0637\u064a\u0632\u202c\u202b\u0650\u0645\u0632\u202c \u202b\u0627\u0626\u202c \u202b\u064e\u0639\u202c \u202b\u0637\u202c \u202b\u0637\u064a\u064c\u0631\u202cBird
Animal\u202b\u0633\u0627\u0627\u0641\u202c \u202b\u064c\u0627\u0636\u0646\u202c \u202b\u0647\u0627\u0621\u202c \u202b\u062a\u062d\u0632\u064a\u202c\u202b\u0633\u064b\u0627\u0627\u0641\u202c \u202b\u064c\u0627\u0631\u0636\u202c \u202b\u064e\u0639\u0627\u0631\u202c \u202b\u062d\u202c \u202b\u0650\u0645\u202c \u202b\u0627\u0644\u062b\u202c \u202b\u0647\u064a\u0627\u0647\u202cRodents Reptiles Aquatic-animals\u202b\u0633\u0648\u0644\u202c \u202b\u0650\u062d\u202c \u202b\u064a\u062a\u0627\u202c \u202b\u0650\u062d\u202c \u202b\u0650\u0645\u202c \u202b\u064e\u0651\u0628\u0627\u062e\u202c \u202b\u064a\u202c \u202b\u0650\u0645\u202c \u202b\u0632\u202c \u202b\u0652\u0644\u202c \u202b\u0650\u0645\u0634\u202c \u202b\u0627\u0644\u0642\u202c\u202b\u0648\u062a\u202c \u202b\u064d\u0629\u202c \u202b\u064e\u0651\u0628\u062d\u202c \u202b\u064a\u202c \u202b\u0650\u0645\u202c \u064c \u202b\u0652\u0644\u202c \u202b\u064e\u0639\u062e\u202c \u202b\u0631\u202c\u202b\u0633\u0648\u0627\u0643\u202c \u202b\u064e\u0639\u202c \u202b\u0623\u202c \u202b\u0650\u062d\u202c \u202b\u0627\u0644\u202c \u202b\u064e\u0651\u062b\u062a\u062a\u202c \u202b\u0650\u062d\u202c \u202b\u0652\u062f\u202c \u202b\u064e\u0651\u062b\u202c \u202b\u0650\u0645\u202c \u202b\u064e\u0651\u0628\u0627\u062e\u202c \u202b\u064a\u202c \u202b\u0650\u0645\u202c \u202b\u064e\u0639\u0627\u0631\u202c \u202b\u062d\u202c \u202b\u064e\u0639\u202c \u202b\u0647\u202cFish Pisces Others-aquatic-
animals
", "num": null, "html": null, "text": "and SubjectField characterize the sense of lexical entry. The Definition determines the meaning of sense. While the Context gives an example of using sense. Regarding the SubjectField it describes the use's domain related to a given sense of a lexical entry. The table below contains some examples of domains available in the Arabic LMF standardized dictionary." }, "TABREF2": { "type_str": "table", "content": "
SubjectField based enrichmentAnalysis based enrichment (exhaustive step)
Animal197
Number ofInsect19
Subject FieldPlant242
Culinary39
Total497
Correct assignment45490
Incorrect assignment4352
Recall91,34 %26 %
Precision98 %63 %
", "num": null, "html": null, "text": "" } } } }