Document Gated Reader for Open-Domain Question Answering

Bingning Wang,Ting Yao,Qi Zhang,Jingfang Xu,Zhixing Tian,Kang Liu,Jun Zhao

DOI: https://doi.org/10.1145/3331184.3331190

2019-01-01

Abstract:Open-domain question answering focuses on using diverse information resources to answer any types of question. Recent years, with the development of large-scale data set and various deep neural networks models, some recent advances in open domain question answering system first utilize the distantly supervised dataset as the knowledge resource, then apply deep learning based machine comprehension techniques to generate the right answers, which achieves impressive results compared with traditional feature-based pipeline methods. However, these deep learning based methods suffer from the inferior quality of the distantly supervised data, and the answer score is un-normalized among multiple documents. Furthermore, unlike previous open-domain question answering system, they process each document independently which may ignore the valuable information in the context. In this paper, we propose a document gated reader to generate the right answer from multiple documents. We propose a document-level gate operation to determine the question-document relevance and embed it into the answer generation process, and optimize it with the global normalization objective. We also develop a bootstrapping based data generation scheme to obtain high-quality training data. Experimental results on several question answering datasets show the advantage of the proposed methods.