BriefGPT.xyz
Oct, 2021
面向开放域段落检索的句子感知对比学习
Representation Decoupling for Open-Domain Passage Retrieval
HTML
PDF
Bohong Wu, Zhuosheng Zhang, Jinyuan Wang, Hai Zhao
TL;DR
本文提出改进内部表示粒度,引入基于上下文句子的模型和 in-passage 负采样策略,以缓解不合适的建模方式对于对话的影响,在三个基准数据集上进行实验并验证方法的有效性,特别是在冲突严重的数据集中的表现。
Abstract
Training dense passage representations via
contrastive learning
(CL) has been shown effective for
open-domain passage retrieval
(ODPR). Recent studies mainly focus on optimizing this CL framework by improving the
→