MONNET · Multilingual Ontologies for Networked Knowledge

The Monnet project will provide a semantics-based solution for integrated information access across language barriers, which is of growing importance to industry as exemplified by the Monnet use cases. A key solution to this problem is to deal with information at the semantic level, i.e. by abstracting away over language and form, allowing for more advanced and uniform: i) integration, ii) aggregation, iii) querying and iv) presentation of information across languages. The Monnet approach therefore consists of: systematic investigation of ontologies as normalizing language-independent semantic models that permit moving flexibly between languages in information extraction, integration and access; provision of automatic, cost-effective localisation of ontologies through innovative approaches exploiting existing web-based resources (lexica, corpora, translation services, etc.); the exploitation of localized ontologies for cross-lingual information extraction from heterogeneous data sources, including text documents, semi-structured tables and increasingly also structured and 'linked data' available in web-accessible RDF knowledge bases; development of flexible multilingual knowledge access and presentation components designed in a principled way based on ontology-lexicon models and multilingual localized ontologies. As outcome, Monnet will have a clear impact by: i) developing models, semi-automatic approaches and methodologies allowing cost effective ontology localization as a basis for ii) implementing an integrated solution to providing semantic-level access to information across languages and iii) providing new evaluation methodologies that evaluate systems in an end-to-end in vivo" fashion to understand user interaction, in addition to more traditional "in vitro" (batch-mode) evaluations for system tuning