Modèle de compréhension du besoin en information pour la RI conversationnelle

Abstract : IR is based on a standard framework that queries document collections through an information need expressed as a set of keywords. Our contribution aims at overpassing this usual paradigm by starting the retrieval process from the natural language expression of the information need; giving rise to a new generation of IR systems based on conversational features (also called "search-oriented conversational systems"). Therefore, the first step focuses on the query formulation from the information need expressed in natural language. In this paper, we propose a query formulation model able of 1) translating natural language expressions to keyword queries in a supervised manner, and 2) injecting relevance feedback in the learning process through reinforcement learning technics. To overcome the lack of training data, we consider the translation model as a word selection process. Our model is evaluated on two TREC collections to demonstrate its effectiveness.
Complete list of metadatas

Cited literature [41 references]  Display  Hide  Download

https://hal.sorbonne-universite.fr/hal-02150653
Contributor : Laure Soulier <>
Submitted on : Friday, June 7, 2019 - 12:48:39 PM
Last modification on : Thursday, October 24, 2019 - 2:44:11 PM

File

CORIA_2019_paper_10.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-02150653, version 1

Citation

Wafa Aissa, Laure Soulier, Ludovic Denoyer. Modèle de compréhension du besoin en information pour la RI conversationnelle. CORIA 2019 - 16ème COnférence en Recherche d’Information et Applications, Mar 2019, Lyon, France. ⟨hal-02150653⟩

Share

Metrics

Record views

53

Files downloads

43