Recherche d'information précise dans des sources d'information structurées et non structurées : défis, approches et hybridation

This paper provides a synthesis of text-based Question Answering (QA) approaches, with emphasis on models exploiting structured representations of texts, and recent QA approaches for knowledge bases. Our goal is to show the common issues, and to identify what can be a common approach, using both the relations from textual documents and the triplets from knowledge bases. We present a few works using this type of approach in order to highlight open issues and pave the way to really hybrid systems based on semantic representations. MOTS-CLÉS : questions-réponses sur des textes, questions-réponses sur bases de connaissances.