Published as a conference paper at ICLR 2019 COARSE-GRAIN FINE-GRAIN COATTENTION NET- WORK FOR MULTI-EVIDENCE QUESTION ANSWERING Victor Zhong1, Caiming Xiong2, Nitish Shirish Keskar2, and Richard Socher2 1Paul G. Allen School of Computer Science & Engineering, University of Washington, Seattle, WA
[email protected] 2Salesforce Research, Palo Alto, CA fcxiong, nkeskar,
[email protected] ABSTRACT End-to-end neural models have made significant progress in question answering, however recent studies show that these models implicitly assume that the answer and evidence appear close together in a single document. In this work, we propose the Coarse-grain Fine-grain Coattention Network (CFC), a new question answer- ing model that combines information from evidence across multiple documents. The CFC consists of a coarse-grain module that interprets documents with respect to the query then finds a relevant answer, and a fine-grain module which scores each candidate answer by comparing its occurrences across all of the documents with the query. We design these modules using hierarchies of coattention and self- attention, which learn to emphasize different parts of the input. On the Qangaroo WikiHop multi-evidence question answering task, the CFC obtains a new state- of-the-art result of 70.6% on the blind test set, outperforming the previous best by 3% accuracy despite not using pretrained contextual encoders. 1 INTRODUCTION A requirement of scalable and practical question answering (QA) systems is the ability to reason over multiple documents and combine their information to answer questions. Although existing datasets enabled the development of effective end-to-end neural question answering systems, they tend to focus on reasoning over localized sections of a single document (Hermann et al., 2015; Rajpurkar et al., 2016; 2018; Trischler et al., 2017).