Enregistré dans:
Détails bibliographiques
Auteurs principaux: Yu, Shi, Fan, Chenghao, Xiong, Chenyan, Jin, David, Liu, Zhiyuan, Liu, Zhenghao
Format: Preprint
Publié: 2023
Sujets:
Accès en ligne:https://arxiv.org/abs/2305.14685
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909188828954624
author Yu, Shi
Fan, Chenghao
Xiong, Chenyan
Jin, David
Liu, Zhiyuan
Liu, Zhenghao
author_facet Yu, Shi
Fan, Chenghao
Xiong, Chenyan
Jin, David
Liu, Zhiyuan
Liu, Zhenghao
contents Common document ranking pipelines in search systems are cascade systems that involve multiple ranking layers to integrate different information step-by-step. In this paper, we propose a novel re-ranker Fusion-in-T5 (FiT5), which integrates text matching information, ranking features, and global document information into one single unified model via templated-based input and global attention. Experiments on passage ranking benchmarks MS MARCO and TREC DL show that FiT5, as one single model, significantly improves ranking performance over complex cascade pipelines. Analysis finds that through attention fusion, FiT5 jointly utilizes various forms of ranking information via gradually attending to related documents and ranking features, and improves the detection of subtle nuances. Our code is open-sourced at https://github.com/OpenMatch/FiT5.
format Preprint
id arxiv_https___arxiv_org_abs_2305_14685
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Fusion-in-T5: Unifying Document Ranking Signals for Improved Information Retrieval
Yu, Shi
Fan, Chenghao
Xiong, Chenyan
Jin, David
Liu, Zhiyuan
Liu, Zhenghao
Information Retrieval
Common document ranking pipelines in search systems are cascade systems that involve multiple ranking layers to integrate different information step-by-step. In this paper, we propose a novel re-ranker Fusion-in-T5 (FiT5), which integrates text matching information, ranking features, and global document information into one single unified model via templated-based input and global attention. Experiments on passage ranking benchmarks MS MARCO and TREC DL show that FiT5, as one single model, significantly improves ranking performance over complex cascade pipelines. Analysis finds that through attention fusion, FiT5 jointly utilizes various forms of ranking information via gradually attending to related documents and ranking features, and improves the detection of subtle nuances. Our code is open-sourced at https://github.com/OpenMatch/FiT5.
title Fusion-in-T5: Unifying Document Ranking Signals for Improved Information Retrieval
topic Information Retrieval
url https://arxiv.org/abs/2305.14685