Teaching Machines to Read and Comprehend
Karl Moritz Hermann† Tomáš Kočiský†‡ Edward Grefenstette† Lasse Espeholt† Will Kay† Mustafa Suleyman† Phil Blunsom†‡ †Google DeepMind ‡University of Oxford
Abstract
Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In this work we define a new methodology that resolves this bottleneck and provides large scale supervised reading comprehension data. This allows us to develop a class of attention based deep neural networks that learn to read real documents and answer complex questions with minimal prior knowledge of language structure.
中文速览
让机器真正读懂一篇文章并回答问题，长期以来缺乏足够大的有标注训练数据，导致这个方向进展缓慢。研究者利用CNN和每日邮报网站上新闻文章与配套摘要要点的天然对应关系，将摘要改写成完形填空（Cloze）式问答，同时对文中实体进行匿名化和随机置换，构建了约100万条"文档—问题—答案"三元组数据集，从根本上解决了数据瓶颈。在此基础上，他们提出了几种基于注意力机制（attention mechanism）的深度循环神经网络模型，其中"专注阅读器（Attentive Reader）"和"迫不及待阅读器（Impatient Reader）"能够在读取问题的同时动态聚焦文档中的相关片段，实验结果显示这些神经网络模型明显优于传统的框架语义分析和基于词距离的基线方法。这项工作的重要意义在于，它同时提供了可公开复现的大规模真实语料和可解释的神经阅读模型，为机器阅读理解研究奠定了重要的数据与方法基础。
原文 arXiv:1506.03340;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1506.03340v3