Knowledge Based Transformer Model for Information Retrieval

Jibril Frej (1) , Jean-Pierre Chevallet (2) , Didier Schwab (3)

1 LIG - Laboratoire d'Informatique de Grenoble
2 MRIM - Modélisation et Recherche d’Information Multimédia [Grenoble]
3 GETALP - Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole

Vocabulary mismatch is a frequent problem in information retrieval (IR). It can occur when the query is short and/or ambiguous but also in specialized domains where queries are made by non-specialists and documents are written by experts. Recently, vocabulary mismatch has been addressed with neural learning-to-rank (NLTR) models and word embeddings to avoid relying only on the exact matching of terms for retrieval. Another approach to vocabulary mismatch is to use knowledge bases (KB) that can associate different terms to the same concept. Given the recent success of transformer encoders for NLP, we propose KTRel: a NLTR model that uses word embeddings, Knowledge bases and Transformer encoders for IR.

Mots clés: Information Retrieval; Neural Networks; Learning-to-Rank; Knowledge Base

Domaines: Recherche d'information [cs.IR]; Traitement du texte et du document; Intelligence artificielle [cs.AI]

Jibril Frej, Jean-Pierre Chevallet, Didier Schwab. Knowledge Based Transformer Model for Information Retrieval. Joint Conference of the Information Retrieval Communities in Europe (CIRCLE 2020), Jul 2020, Samatan, France. ⟨hal-03263784⟩