arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.06499cs.IR
收藏

基于BERT的复合重排序算法

Composite Re-Ranking for Efficient Document Search with BERT

Yingrui Yang, Yifan Qiao, Jinjin Shao, Mayuresh Anand, Xifeng Yan, Tao Yang

摘要:

虽然相当大的努力已经投入到基于变压器的排名模型的文档搜索,相关性效率的权衡仍然是一个关键问题的特设排名。为了克服这一挑战,本文提出了BECR(基于BERT的复合重新排序),一个复合重新排序方案,结合了深度上下文令牌交互和传统的词汇术语匹配功能。特别地,BECR利用令牌编码机制将查询表示分解为可预计算的uni-gram和skip-n-gram。通过在双编码器架构之上应用令牌编码,BECR在捕获查询的上下文语义的同时分离了查询和文档之间的注意力。与以前的方法相比,该框架在在线推理期间不执行昂贵的BERT计算。因此,它是显着更快,但仍然能够实现高竞争力的ad-hoc排名的相关性。最后,使用TREC数据集描述了BECR和其他最先进的神经排名基线之间的深入比较,从而进一步证明了BECR的增强的相关性和效率。

英文摘要:

Although considerable efforts have been devoted to transformer-based ranking models for document search, the relevance-efficiency tradeoff remains a critical problem for ad-hoc ranking. To overcome this challenge, this paper presents BECR (BERT-based Composite Re-Ranking), a composite re-ranking scheme that combines deep contextual token interactions and traditional lexical term-matching features. In particular, BECR exploits a token encoding mechanism to decompose the query representations into pre-computable uni-grams and skip-n-grams. By applying token encoding on top of a dual-encoder architecture, BECR separates the attentions between a query and a document while capturing the contextual semantics of a query. In contrast to previous approaches, this framework does not perform expensive BERT computations during online inference. Thus, it is significantly faster, yet still able to achieve high competitiveness in ad-hoc ranking relevance. Finally, an in-depth comparison between BECR and other start-of-the-art neural ranking baselines is described using the TREC datasets, thereby further demonstrating the enhanced relevance and efficiency of BECR.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑