From Dictionaries to Cross-lingual Lexical Resources Guadalupe - PowerPoint PPT Presentation

From Dictionaries to Cross-lingual Lexical Resources Guadalupe Aguado-de-Cea, Elena Montiel-Ponsoda, Ilan Kernerman, Noam Ordan Ontology Engineering Group (OEG) Universidad Politécnica de Madrid (UPM) {lupe,emontiel}@fi.upm.es K Dictionaries, Tel Aviv {ilan, noam}@kdictionaries.com Acknowledgements Spported by the Spanish Ministry of Economy and Competitiveness project 4V (TIN2013-46238-C4-2-R), the Excellence Network ReTeLe (TIN2015-68955-REDT), and the Juan de la Cierva program and by the Spanish Julia Bosque-Gil Ministry of Education, Culture and Sports through the FPU program.

WHY MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

Data in proprietary formats: • Isolated • Not interoperable • Multiple access points • Duplicities MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

http://es.wiktionary.org http://rae.es “Red” (comp uter network) http://www.wikilengua.org/ index.php/Terminesp:red http:/www.apertium.org http://kdictionaries.com http://es.wikipedia.org MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

“ Red ” Pronunciation: [red] Grammar category: sustantivo femenino “ Red ” Singular: “ red ” Etimology: Del latin “ rete ” Plural: “ redes ” Gender: “ f ” Definition: “ Conjunto de “ Red ” ordenadores o de equipos informáticos conectados entre Complementary Norm: UNE 21302-131 sí … . ” English: network information, but German: Netzwerk not linked “ Red ” translations: “ xarxa ” (ca), “ rede ” (ga), … “ Red_de_computadores ” “ Red ” Category: redes informáticas Grammatical categoty: n, f Image Definition: “ conjunto de computadoras con (...) translation: “ rede ” (pt) *Picture attribution: http://commons.wikimedia.org/wiki/User:Gugerell MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

Linguistic Linked Open Data cloud NIF NLP Interchange Format MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

WHAT MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

K Dictionaries multi-language Global Series • Spanish set of the K Dictionaries (KD) multi-language Global Series (24 languages) • The approach followed in this series is to compile for each language a core vocabulary as a standalone project and have it translated to other languages in more projects. • Not biased towards any language, and each is represented on its own terms. • Translated to another language at a later phase, creating a pair-specific, and thus pair- sensitive, interlingual representation http://kdictionaries-online.com/ MetaForum 2016, 4-5 .07.16, Lisboa, Portugal 4

K Dictionaries XML proprietary format Multilingual information for the Spanish headword acalorado (heated) <SenseGrp identifier="SE00000730" version="1"> <Synonym>agitado</Synonym> <Definition>que es muy animado</Definition> <TranslationCluster identifier="TC00001663" text="que es muy animado" type="def"> <Locale lang=" nl "> Dutch <TranslationBlock> <TranslationCtn> <Translation>verhit</Translation> </TranslationCtn> <TranslationCtn> <Translation>vurig</Translation> </TranslationCtn> </TranslationBlock> </Locale> Norwegian <Locale lang=" no "> <TranslationBlock> <TranslationCtn> <Translation>ivrig, oppsatt, opphetet</Translation> </TranslationCtn> </TranslationBlock> [ … ] 4

K Dictionaries XML proprietary format [ … ] </TranslationCluster> <ExampleCtn type="sid" version="1"> <Example>sesión acalorada</Example> <TranslationCluster identifier="TC00001664" text="sesión acalorada" type="exmp"> <Locale lang="nl"> <TranslationBlock> <TranslationCtn> <Translation>vurige zitting</Translation> </TranslationCtn> </TranslationBlock> </Locale> <Locale lang="no"> <TranslationBlock> <TranslationCtn> <Translation>et opphetet møte</Translation> </TranslationCtn> </TranslationBlock> [ … ] </TranslationCluster> 4

HOW MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

5 tasks to migrate resources into linked data (Vila Suero et al., 2014): 1. Data exploration 2. URI naming strategy definition 3. Modeling 4. RDF generation 5. Linking MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

5 tasks to migrate resources into linked data (Vila Suero et al., 2014): 1. Data exploration 2. URI naming strategy definition 3. Modeling 4. RDF generation 5. Linking MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

Modeling lemon-ontolex Model for representing linguistic information in RDF(s) (semantics by reference) “An ontology-based semantic lexicon would leave the semantics to the ontology, focusing instead on providing domain-specific terms and object descriptions in the ontology.” (Buitelaar, 2010) • Concise and descriptive (external repositories of linguistic categories) • Modular and extensible (5 modules) MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

The origins … LMF -Lexical Markup Framework (ISO 24613) XML LexInfo, LIR Represent lexical information relative to an ontology OWL SKOS (W3C Standard) Designed for Taxonomy/Vocabulary representation in RDF Nowadays … de facto standard for transforming linguistic resources into the linked data format http://linghub.lider-project.eu/ From McCrae, J. , “ lemon: The Lexicon Model for Ontologies” SD-LLOD-15 MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

lemon-ontolex MetaForum 2016, 4-5 .07.16, Lisboa, Portugal 8

The vartrans module MetaForum 2016, 4-5 .07.16, Lisboa, Portugal 8

Modelling example for acalorado MetaForum 2016, 4-5 .07.16, Lisboa, Portugal 8

CONCLUSIONS MetaForum 2016, 4-5 .07.16, Lisboa, Portugal

LD paradigm • Innovative way of representing and linking data • Keeping language specificities EN DE as much as possible, in a cross- lingual graph, bottom-up fashion • Retrieving language information by linking translations ES IT • In the KD case, RDF links contribute to automatic growth of lexical resources, at a different pace FR • “One-stop shop”, rather than many shops MetaForum 2016, 4-5 .07.16, Lisboa, Portugal 4

Thank you! See you at the poster session … MetaForum 2016, 4-5 .07.16, Lisboa, Portugal