Reading Wikipedia to Answer Open-Domain Questions

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1870–1879 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1171 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1870–1879 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1171 Reading Wikipedia to Answer Open-Domain Questions Danqi Chen∗ Computer Science Stanford University Stanford, CA 94305, USA danqi@cs.stanford.edu Adam Fisch, Jason Weston & Antoine Bordes Facebook AI Research 770 Broadway New York, NY 10003, USA {afisch,jase,abordes}@fb.com Abstract This paper proposes to tackle open- domain question answering using Wikipedia as the unique knowledge source: the answer to any factoid question is a text span in a Wikipedia article. This task of machine reading at scale combines the challenges of document re- trieval (finding the relevant articles) with that of machine comprehension of text (identifying the answer spans from those articles). Our approach combines a search component based on bigram hashing and TF-IDF matching with a multi-layer recurrent neural network model trained to detect answers in Wikipedia paragraphs. Our experiments on multiple existing QA datasets indicate that (1) both modules are highly competitive with respect to existing counterparts and (2) multitask learning using distant supervision on their combination is an effective complete system on this challenging task. 1 Introduction This paper considers the problem of answering factoid questions in an open-domain setting us- ing Wikipedia as the unique knowledge source, such as one does when looking for answers in an encyclopedia. Wikipedia is a constantly evolv- ing source of detailed information that could fa- cilitate intelligent machines — if they are able to leverage its power. Unlike knowledge bases (KBs) such as Freebase (Bollacker et al., 2008) or DB- Pedia (Auer et al., 2007), which are easier for computers to process but too sparsely populated for open-domain question answering (Miller et al., ∗ Most of this work was done while DC was with Face- book AI Research. 2016), Wikipedia contains up-to-date knowledge that humans are interested in. It is designed, how- ever, for humans – not machines – to read. Using Wikipedia articles as the knowledge source causes the task of question answering (QA) to combine the challenges of both large-scale open-domain QA and of machine comprehension of text. In order to answer any question, one must first retrieve the few relevant articles among more than 5 million items, and then scan them care- fully to identify the answer. We term this setting, machine reading at scale (MRS). Our work treats Wikipedia as a collection of articles and does not rely on its internal graph structure. As a result, our approach is generic and could be switched to other collections of documents, books, or even daily up- dated newspapers. Large-scale QA systems like IBM’s DeepQA (Ferrucci et al., 2010) rely on multiple sources to answer: besides Wikipedia, it is also paired with KBs, dictionaries, and even news articles, books, etc. As a result, such systems heavily rely …

Content truncated to 500 words