Frameworks, Implementation & Open Frameworks, Implementation & Open Problems for the Collaborative Building of Problems for the Collaborative Building of a Multilingual Lexical Database a Multilingual Lexical Database Mathieu Mangeot & Gilles Sérasset NII, Tokyo, GETA-CLIPS, Grenoble, Japan France mangeot@nii.ac.jp Gilles.Serasset@imag.fr Sat. 31 Aug. 2002 SEMANET Workshop 1 / 25
Outline Outline ♦ Presentation of Papillon Project ♦ Macrostructure of the Dictionary ♦ Microstructure of the Entries ♦ Bootstrapping & Contribution Process ♦ Limbo, Purgatory & Paradise ♦ Bootstrapping with Conceptual Vectors ♦ Contributions & Validation Process ♦ Lexico-Semantical Network ♦ Monolingual with Lexical Functions ♦ Multilingual with Axies (Interlingual Links) ♦ Conclusion & References Sat. 31 Aug. 2002 SEMANET Workshop 2 / 25
Motivations Motivations ♦ Initial Goal ♦ Build a French-Japanese electronic dictionary for humans ♦ Lack of Information ♦ Numerical Specifiers, kanji+kana+romaji ♦ Very Few Existing Resources ♦ French-Japanese, Free, Electronic ♦ Construction Costs Too High ♦ EDR English-Japanese Dictionary ♦ 1200 human-year; 300 000 entries; price: 14,3 Mo ¥ ♦ On Going Collaborative Construction Projects ♦ Edict Japanese->English, SAIKAM Japanese-Thai Sat. 31 Aug. 2002 SEMANET Workshop 3 / 25
Extended Goals Extended Goals ♦ Build a More Complete Dictionary ♦ Multilingual (English, French, German, Japanese, Lao , Malay, Thai, Vietnamese) ♦ Multiusers (beginners, experts, applications) ♦ Community Development ♦ LINUX Construction Paradigm ♦ Voluntary Contributors ♦ Mutualization of the Resources ♦ User Preferences & Profiles Sat. 31 Aug. 2002 SEMANET Workshop 4 / 25
Outline Outline ♦ Presentation of Papillon Project ♦ Macrostructure of the Dictionary ♦ Microstructure of the Entries ♦ Bootstrapping & Contribution Process ♦ Limbo, Purgatory & Paradise ♦ Bootstrapping with Conceptual Vectors ♦ Contributions & Validation Process ♦ Lexico-Semantical Network ♦ Monolingual with Lexical Functions ♦ Multilingual with Axies (Interlingual Links) ♦ Conclusion & References Sat. 31 Aug. 2002 SEMANET Workshop 5 / 25
Bilingual Dictionaries Bilingual Dictionaries French Vietnamese Japanese Thai English Lao Malay Sat. 31 Aug. 2002 SEMANET Workshop 6 / 25
Pivot Dictionary Pivot Dictionary French Vietnamese Japanese Thai Int English Lao Malay Sat. 31 Aug. 2002 SEMANET Workshop 7 / 25
Detailed Pivot Structure Detailed Pivot Structure French DiCo English DiCo Vocable affection N Vocable affection n.f. Interlingual Links (Axies) lexie affection.1 lexie affection (tendresse) Vocable disease N lexie affection.2 lexie disease (médecine) Vocable maladie n.f. lexie maladie Japanese DiCo 病気 byouki 【びょうき】 Ref: Work done by Gilles Sérasset Refinement Links Sat. 31 Aug. 2002 SEMANET Workshop 8 / 25
Outline Outline ♦ Presentation of Papillon Project ♦ Macrostructure of the Dictionary ♦ Microstructure of the Entries ♦ Bootstrapping & Contribution Process ♦ Limbo, Purgatory & Paradise ♦ Bootstrapping with Conceptual Vectors ♦ Contributions & Validation Process ♦ Lexico-Semantical Network ♦ Monolingual with Lexical Functions ♦ Multilingual with Axies (Interlingual Links) ♦ Conclusion & References Sat. 31 Aug. 2002 SEMANET Workshop 9 / 25
Combinatorial Lexicography Combinatorial Lexicography ♦ From Meaning-Text Theory ♦ Alain Polguère & Igor Mel ’ tchuk (U. de Montréal) ♦ Gives the necessary information to go from an idea (the meaning) to its realisation in a given language (the text). ♦ Existing Dictionaries: DEC, DiCo database & LAF ♦ Same Structure for Every Language ♦ 56 Basic Lexical Functions Sat. 31 Aug. 2002 SEMANET Workshop 10 / 25
French Lexie (DiCo Entry) French Lexie (DiCo Entry) Name of the Lexical Unit: MEURTRE ♦ Grammatical Properties: nom, masc ♦ Semantical Formula: action de tuer: ~ PAR L'individu X DE ♦ L'individu Y Government Pattern: X =I = de N, A-poss Y= II = de N, A-poss ♦ Lexical Functions: ♦ {QSyn} assassinat,homicide#1;crime /*Quasi synonyms*/ ♦ {Oper 1 } accomplir, commettre, perpétrer [ART ~]; ♦ tremper [dans ART ~] /*Causes that X does a M.*/ {S 1 } auteur [de ART Ø]//meurtrier-n /*Name for X*/ ♦ {S 2 } victime [de ART Ø] /*Name for Y*/ ♦ Example: La mésentente pourrait être le mobile du meurtre. ♦ Idioms: ♦ _appel au meurtre_ ♦ _crier au meurtre_ ♦ Sat. 31 Aug. 2002 SEMANET Workshop 11 / 25
【え】るだろう。 Japanese Lexie Japanese Lexie Name of the Lexical Unit: 殺人 【さつじん】 ♦ Reading: satsujin ♦ Grammatical Properties: 名詞 【めいし】 ♦ Semantical Formula: どうさ : 人 Y の 人 X の ̃ ♦ Government Pattern: X = I = N, Y = II = N の ♦ Lexical Functions: ♦ {QSyn} 殺戮【さつりく】 , 殺害【さつがい】 /*Quasi synonyms*/ ♦ {Oper 1 } [̃ を ] する ； [̃ を ] 犯す /* Causes that X does a M. */ ♦ {S 1 } 殺人者 【さつじんしゃ】 , 殺人鬼 【さつじんき】 /*Name for X*/ ♦ {S 2 } 被害者【ひがいしゃ】 /*Name for Y*/ ♦ Example: 喧嘩【けんか】は殺人【さつじん】の動機【どうき】になり得 ♦ Idioms: ♦ _ 殺人剣 【さつじんけん】 _ ♦ _ 嘱託殺人 【しくたくさつじん】 _ ♦ Sat. 31 Aug. 2002 SEMANET Workshop 12 / 25
Interlingual Links (Axies) Interlingual Links (Axies) ♦ Linking Monolingual Lexies ♦ Motivated by existing translation links. ♦ Not like concepts ♦ Links to other Axies ♦ Synonyms, Refinement, Generalizations ♦ Links to External References ♦ To be independent from any existing theory ♦ Wordnet synsets, NTT Semantic category, ♦ ONTOS or LexiGuide ontologies, ♦ UNL Uws & Graphs etc. Sat. 31 Aug. 2002 SEMANET Workshop 13 / 25
Structure of an Axie Structure of an Axie Unique ID: a000023 ♦ Semantic Tag (entity, process, state, result): process ♦ Links to lexies: fra: meurtre.1 eng: murder.1 jpn: satsujin.1 ♦ Links to other axies ♦ ♦ synonym axies: a000024 (assassination) ♦ generic axies: a00002 ♦ refined axies: a000025 References to External Resources: ♦ ♦ WordNet Synset: 00143589 unlawful premeditated killing of a human being ♦ UNL UW: murder(icl>action,agt>human,obj>human) ♦ NTT Semantic Category ♦ ONTOS Concept ♦ LexiGuide concept Sat. 31 Aug. 2002 SEMANET Workshop 14 / 25
Outline Outline ♦ Presentation of Papillon Project ♦ Macrostructure of the Dictionary ♦ Microstructure of the Entries ♦ Bootstrapping & Contribution Process ♦ Limbo, Purgatory & Paradise ♦ Bootstrapping with Conceptual Vectors ♦ Contributions & Validation Process ♦ Lexico-Semantical Network ♦ Monolingual with Lexical Functions ♦ Multilingual with Axies (Interlingual Links) ♦ Conclusion & References Sat. 31 Aug. 2002 SEMANET Workshop 15 / 25
Preparation of the Existing Data Preparation of the Existing Data Limbo Local Resources Original Format DicDist Purgatory SAIKAM FeM Consultation XML Format I Recuperation n JMDict t e g r a t Paradise i o n Papillon Format DiCo S pap DicOrig Contrib1 Import EDR Integration Contrib2 ELRA Contrib3 Export DicGen Sat. 31 Aug. 2002 SEMANET Workshop 16 / 25
Introduction to Conceptual Vectors Introduction to Conceptual Vectors ♦ An idea = a concept = a conceptual vector ♦ The vector space is of K dimensions ♦ K = nb of concepts in a thesaurus hierarchy ♦ Eg: for French, Thesaurus Larousse = 873 concepts ♦ One independent vector space for each language ♦ Distance between 2 vectors = angular distance ♦ D A (x, y) = acos(sim(x,y)) ♦ D A (x, y) = acos(x.y/|x||y|)) ♦ Ref: Work done by Mathieu Lafourcade ♦ http://www.lirmm.fr/~lafourca/ Sat. 31 Aug. 2002 SEMANET Workshop 17 / 25
Linking Word Senses with Vectors Linking Word Senses with Vectors English-French Vector space English Vectorized Bilingual dictionary for English monolingual dictionary demand demand equivalents demand.1 v demand.1 def v equivalents demand.2 v demand.2 def v v equivalents demand.3 demand.3 def v v equivalents demand.4 Association Left over meaning Slide from Mathieu Lafourcade Sat. 31 Aug. 2002 SEMANET Workshop 18 / 25
Contributions & Validation Contributions & Validation Papillon Server Paradise Voluntary Payed Papillon Format Contributors Specialists User Space Validation S pap Contributions Integration User Space Sat. 31 Aug. 2002 SEMANET Workshop 19 / 25
Outline Outline ♦ Presentation of Papillon Project ♦ Macrostructure of the Dictionary ♦ Microstructure of the Entries ♦ Bootstrapping & Contribution Process ♦ Limbo, Purgatory & Paradise ♦ Bootstrapping with Conceptual Vectors ♦ Contributions & Validation Process ♦ Lexico-Semantical Network ♦ Monolingual with Lexical Functions ♦ Multilingual with Axies (Interlingual Links) ♦ Conclusion & References Sat. 31 Aug. 2002 SEMANET Workshop 20 / 25
Lexico Semantical Lexico Semantical Multilingual Network (1) Multilingual Network (1) Interlingual Links DiCo English DiCo French Assassinat Assassination _To call for sb's {Qsyn} ? assassination_ {Qsyn} {Qsyn} Meurtre Murder _Lancer un appel au meurtre_ Sat. 31 Aug. 2002 SEMANET Workshop 21 / 25
Recommend
More recommend
