CoSPLADE: Contextualizing SPLADE for Conversational Information Retrieval

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hai, Nam Le, Gerald, Thomas, Formal, Thibault, Nie, Jian-Yun, Piwowarski, Benjamin, Soulier, Laure
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917713119543296
author Hai, Nam Le
Gerald, Thomas
Formal, Thibault
Nie, Jian-Yun
Piwowarski, Benjamin
Soulier, Laure
author_facet Hai, Nam Le
Gerald, Thomas
Formal, Thibault
Nie, Jian-Yun
Piwowarski, Benjamin
Soulier, Laure
contents Conversational search is a difficult task as it aims at retrieving documents based not only on the current user query but also on the full conversation history. Most of the previous methods have focused on a multi-stage ranking approach relying on query reformulation, a critical intermediate step that might lead to a sub-optimal retrieval. Other approaches have tried to use a fully neural IR first-stage, but are either zero-shot or rely on full learning-to-rank based on a dataset with pseudo-labels. In this work, leveraging the CANARD dataset, we propose an innovative lightweight learning technique to train a first-stage ranker based on SPLADE. By relying on SPLADE sparse representations, we show that, when combined with a second-stage ranker based on T5Mono, the results are competitive on the TREC CAsT 2020 and 2021 tracks.
format Preprint
id arxiv_https___arxiv_org_abs_2301_04413
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle CoSPLADE: Contextualizing SPLADE for Conversational Information Retrieval
Hai, Nam Le
Gerald, Thomas
Formal, Thibault
Nie, Jian-Yun
Piwowarski, Benjamin
Soulier, Laure
Information Retrieval
Conversational search is a difficult task as it aims at retrieving documents based not only on the current user query but also on the full conversation history. Most of the previous methods have focused on a multi-stage ranking approach relying on query reformulation, a critical intermediate step that might lead to a sub-optimal retrieval. Other approaches have tried to use a fully neural IR first-stage, but are either zero-shot or rely on full learning-to-rank based on a dataset with pseudo-labels. In this work, leveraging the CANARD dataset, we propose an innovative lightweight learning technique to train a first-stage ranker based on SPLADE. By relying on SPLADE sparse representations, we show that, when combined with a second-stage ranker based on T5Mono, the results are competitive on the TREC CAsT 2020 and 2021 tracks.
title CoSPLADE: Contextualizing SPLADE for Conversational Information Retrieval
topic Information Retrieval
url https://arxiv.org/abs/2301.04413