Question Answering with Subgraph Embeddings

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 615–620, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Question Answering with Subgraph Embeddings Antoine Bordes Facebook AI Research 112 avenue de Wagram, Paris, France abordes@fb.com Sumit Chopra Facebook AI Research 770 Broadway, New York, USA spchopra@fb.com Jason Weston Facebook AI Research 770 Broadway, New York, USA jase@fb.com Abstract This paper presents a system which learns to answer questions on a broad range of topics from a knowledge base using few hand-crafted features. Our model learns low-dimensional embeddings of words and knowledge base constituents; these representations are used to score natural language questions against candidate an- swers. Training our system using pairs of questions and structured representations of their answers, and pairs of question para- phrases, yields competitive results on a re- cent benchmark of the literature. 1 Introduction Teaching machines how to automatically answer questions asked in natural language on any topic or in any domain has always been a long stand- ing goal in Artificial Intelligence. With the rise of large scale structured knowledge bases (KBs), this problem, known as open-domain question an- swering (or open QA), boils down to being able to query efficiently such databases with natural language. These KBs, such as FREEBASE (Bol- lacker et al., 2008) encompass huge ever growing amounts of information and ease open QA by or- ganizing a great variety of answers in a structured format. However, the scale and the difficulty for machines to interpret natural language still makes this task a challenging problem. The state-of-the-art techniques in open QA can be classified into two main classes, namely, infor- mation retrieval based and semantic parsing based. Information retrieval systems first retrieve a broad set of candidate answers by querying the search API of KBs with a transformation of the ques- tion into a valid query and then use fine-grained detection heuristics to identify the exact answer (Kolomiyets and Moens, 2011; Unger et al., 2012; Yao and Van Durme, 2014). On the other hand, semantic parsing methods focus on the correct in- terpretation of the meaning of a question by a se- mantic parsing system. A correct interpretation converts a question into the exact database query that returns the correct answer. Interestingly, re- cent works (Berant et al., 2013; Kwiatkowski et al., 2013; Berant and Liang, 2014; Fader et al., 2014) have shown that such systems can be ef- ficiently trained under indirect and imperfect su- pervision and hence scale to large-scale regimes, while bypassing most of the annotation costs. Yet, even if both kinds of system have shown the ability to handle large-scale KBs, they still require experts to hand-craft lexicons, grammars, and KB schema to be effective. This non-negligible hu- man intervention might not be generic enough to conveniently scale up to new databases with other schema, broader vocabularies or languages other than English. In contrast, (Fader et al., 2013) pro- posed a framework for open QA requiring almost no human annotation. Despite …

Content truncated to 500 words