摘要
该文介绍THUIR团队在"2018机器阅读理解技术竞赛"中的模型设计与实验结果。针对多文档机器阅读理解任务,设计了基于自注意力机制的多任务深度阅读理解模型T-Reader,在所有105支参赛队伍中取得了第八名的成绩。除文本信息外,提取了问题与段落精准匹配等特征作为模型输入;在模型的段落匹配阶段,采用跨段落的文档级自注意力机制,通过循环神经网络实现了跨文档的问题级信息交互;在答案范围预测阶段,通过进行段落排序引入强化学习的方法提升模型性能。
This paper describes the approach and the experimental results of THUIR at 2018 NLP Challenge on Machine Reading Comprehension.We design a multi-task deep neural model with self-attention mechanism.The selfattention mechanism on passages within a document allows information to flow across passages,and the recurrent neural network further shares information across documents.Besides the distributed representation of questions and passages learned during model training,we also extract features denoting exact matching between questions and passages as the inputs of the model.When predicting the span of the answer,we introduce passage ranking into the model to promote the model performance via reinforcement learning.This proposed method ranks 8/105 on the final test set.
作者
郑玉昆
李丹
范臻
刘奕群
张敏
马少平
ZHENG Yukun;LI Dan;FAN Zhen;LIU Yiqun;ZHANG Min;MA Shaoping(Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China;University of Amsterdafn ILPS,Amsterdam 1098XH,Netherlands)
出处
《中文信息学报》
CSCD
北大核心
2018年第11期128-134,共7页
Journal of Chinese Information Processing
基金
国家自然科学基金(61622208
61732008
61532011)
国家973计划(2015CB358700)
关键词
机器阅读理解
问答系统
深度学习
强化学习
machine reading comprehension
question answering system
deep learning
reinforcement learning
