摘要
机器阅读理解任务要求机器根据篇章文本回答相关问题。该文以抽取式机器阅读理解为例,重点考察当问题的线索要素与答案在篇章文本中跨越多个标点句时的阅读理解问题。该文将小句复合体结构自动分析任务与机器阅读理解任务融合,利用小句复合体中跨标点句话头-话体共享关系,来降低机器阅读理解任务的难度;并设计与实现了基于小句复合体的机器阅读理解模型。实验结果表明,在问题线索要素与答案跨越多个标点句时,答案抽取的精确匹配率(EM)相对于基准模型提升了3.49%,模型整体的精确匹配率提升了3.26%。
The machine reading comprehension task requires the machine to answer relevant questions according to the context.Focused on extractive machine reading comprehension,this paper proposes a clause complex based machine reading comprehension method.The naming structure relationship of clause complex is introduced to alleviate the difficult cases with clue elements or answer elements spreading across multiple punctuation sentences.The experimental results show that the proposed method improves the EM of the whole model by 3.26%,and that for the difficult cases by 3.49%.
作者
王瑞琦
罗智勇
刘祥
韩瑞昉
李舒馨
WANG Ruiqi;LUO Zhiyong;LIU Xiang;HAN Ruifang;LI Shuxin(School of Information Science,Beijing Language and Culture University,Beijing 100083,China)
出处
《中文信息学报》
CSCD
北大核心
2024年第3期130-140,共11页
Journal of Chinese Information Processing
基金
国家自然科学基金(62076037)。
关键词
机器阅读理解
跨标点句问答
小句复合体
machine reading comprehension
cross-punctuation
questions and answers
clause complex