Extending the Semantic Layer of the CompL-it Italian Lexicon: Traits, Semantic Types, and Definitions

Abstract

The growing impact of Large Language Models has highlighted the need for explicit, interpretable linguistic knowledge. Lexical resources respond to this need by offering structured representations that complement and constrain the implicit semantics of neural models. This paper presents an extension of CompL-it, currently the most comprehensive open computational lexicon of Italian. Building on the semantic layer inherited from LexicO—itself derived from the PAROLE-SIMPLE-CLIPS resource—the work enriches CompL-it with semantic traits and references to semantic types. Moreover, an experiment was conducted to generate missing definitions through an automatic process supported by LLMs. The resulting resource thus combines human-curated and machine-extended knowledge, ensuring both linguistic precision and scalability. This enriched semantic layer enhances CompL-it’s interoperability within the Linguistic Linked Data framework and strengthens its usability for NLP tasks such as word sense disambiguation, semantic role labelling, and knowledge grounding.

Resources

Details