a friendly localized platform for multilingual semantic communication

Softcust : A Friendly Localized Platform for Multilingual Semantic Communication

Fabian Cretton, Zhan Liu and Anne Le Calvé Institute of Business Information Systems

University of Applied Sciences Western Switzerland Sierre, Switzerland

fabian.cretton@hevs.ch, zhan.liu@hevs.ch anne.lecalve@hevs.ch

04.2013

HES-SO Valais – Sierre - Suisse

OVERVIEW

•  Motivation

•  Objectives of our solution

•  Design & development

•  Demonstration

•  Practical implication

•  Conclusion

MOTIVATION

Ming Alice

Bonjour 你好

Gang

Automatic matching, not just translation

Needs

•  Platform to help users match their competencies, needs, offers and activities –  Natural Language Processing –  Mother tong –  Cross Language

•  Technical needs –  Automatic matching platform –  Full text search –  Language's lexical semantic (synonym, hyponym, …) –  At least 3 languages : French, English, Chinese –  Computer sciences domain –  Easily extensible

RESEARCH QUESTION

How can semantic web technologies improve the performance of an exchange platform in designing a friendly localized

system with multilingual semantic communication?

3. Hybrid search : semantic and classical (for words not in ontology)

OBJECTIVES OF OUR SOLUTION

2 tools for text analysis and search

Stored in RDBMS

1.  Users enter their informations in differents languages

2. Sent to semantic engine for indexing Both semantic and classical solution

DESIGN & DEVELOPMENT Classical Text Search Engine

•  Text Search Engine - Apache Lucene: open source, high-performance, full-featured text search-engine library

DESIGN & DEVELOPMENT Multilingual database

•  Wikipedia Database: multilingual, web-based, freely available knowledge –  Java-based Wikipedia Library (JWPL) –  Extracting lexical semantic knowledge

from Wikipedia –  21 Millions articles, 280 languages –  Three different types of information :

•  Synonyms, •  Hyponyms ("type-of"), •  Meronyms ("part-of")

DESIGN & DEVELOPMENT Multilingual Semantic Search Engine

WikiOnto : SKOS ontology each term : 3 languages + a link to Synonyms, Hyponymy and Meronmy.

DEMONSTRATION Page: Me

User enter their profile in mother tong

DEMONSTRATION Page: My Domain

User describe their domain of activities and upload documents to add more information on their products

DEMONSTRATION Page: My Wish List

Criteria's search for partners : keywords, roles, domain types, countrie and languages.

Results given in the mother tongue of the user

The semantic matching allows cross languages searches.

PRACTICAL IMPLICATION

CONCLUSION

•  Multilingual semantic communication for automated information management through a friendly localized platform for international communications and business models

•  New ontologies, which support the exchange multilingual semantically enriched messages.

•  Hybrid search : semantic and classical

•  Advantages: •  Add More languages (Deutsh, Italian,…) •  Connect to Linked Open Data (wordnet, Geonames, …) •  Applied to other domains (eGov, BPMN,…)

•  Thematic graph of texts

Merci de votre attention