Learning Structured Natural Language Representations for Semantic Parsing

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 44–55 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1005 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 44–55 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1005 Learning Structured Natural Language Representations for Semantic Parsing Jianpeng Cheng† Siva Reddy† Vijay Saraswat‡ and Mirella Lapata† †School of Informatics, University of Edinburgh ‡IBM T.J. Watson Research {jianpeng.cheng,siva.reddy}@ed.ac.uk, vsaraswa@us.ibm.com, mlap@inf.ed.ac.uk Abstract We introduce a neural semantic parser which is interpretable and scalable. Our model converts natural language utter- ances to intermediate, domain-general nat- ural language representations in the form of predicate-argument structures, which are induced with a transition system and subsequently mapped to target domains. The semantic parser is trained end-to-end using annotated logical forms or their de- notations. We achieve the state of the art on SPADES and GRAPHQUESTIONS and obtain competitive results on GEO- QUERY and WEBQUESTIONS. The in- duced predicate-argument structures shed light on the types of representations useful for semantic parsing and how these are dif- ferent from linguistically motivated ones.1 1 Introduction Semantic parsing is the task of mapping natu- ral language utterances to machine interpretable meaning representations. Despite differences in the choice of meaning representation and model structure, most existing work conceptualizes se- mantic parsing following two main approaches. Under the first approach, an utterance is parsed and grounded to a meaning representation directly via learning a task-specific grammar (Zelle and Mooney, 1996; Zettlemoyer and Collins, 2005; Wong and Mooney, 2006; Kwiatkowksi et al., 2010; Liang et al., 2011; Berant et al., 2013; Flanigan et al., 2014; Pasupat and Liang, 2015; Groschwitz et al., 2015). Under the second ap- proach, the utterance is first parsed to an inter- mediate task-independent representation tied to a syntactic parser and then mapped to a grounded 1Our code is available at https://github.com/ cheng6076/scanner. representation (Kwiatkowski et al., 2013; Reddy et al., 2016, 2014; Krishnamurthy and Mitchell, 2015; Gardner and Krishnamurthy, 2017). A merit of the two-stage approach is that it creates reusable intermediate interpretations, which potentially en- ables the handling of unseen words and knowledge transfer across domains (Bender et al., 2015). The successful application of encoder-decoder models (Bahdanau et al., 2015; Sutskever et al., 2014) to a variety of NLP tasks has provided strong impetus to treat semantic parsing as a se- quence transduction problem where an utterance is mapped to a target meaning representation in string format (Dong and Lapata, 2016; Jia and Liang, 2016; Koˇcisk´y et al., 2016). Such models still fall under the first approach, however, in con- trast to previous work (Zelle and Mooney, 1996; Zettlemoyer and Collins, 2005; Liang et al., 2011) they reduce the need for domain-specific assump- tions, grammar learning, and more generally ex- tensive feature engineering. But this modeling flexibility comes at a cost since it is no longer pos- sible to interpret how meaning composition is per- formed. Such knowledge plays a …

Content truncated to 500 words