Experiences with a multilingual ontology-based lexicon for news filtering

Ontologies as part of a lexicon are important for many NLP tasks. We draw on experiences gained in an ESPRIT project called TREVI, which concerns news filtering and enrichment and includes an English/Spanish multilingual lexicon. We sketch the way the lexicon, including an ontology, is constructed and present the methodology used for adding domain ontologies. In general, the approach is lexicon-driven in the sense that ontology (semantics) and lexicon (word forms) are developed in tandem