Towards Comparability of Linguistic Graph Banks for Semantic Parsing

Towards Comparability of Linguistic Graph Banks for Semantic Parsing Stephan Oepen♣, Marco Kuhlmann♠, Yusuke Miyao♥, Daniel Zeman♦, Silvie Cinková♦, Dan Flickinger◦, Jan Hajiˇc♦, Angelina Ivanova♣, and Zdeˇnka Urešová♦ ♣ University of Oslo, Department of Informatics ♠ Linköping University, Department of Computer and Information Science ♥ National Institute of Informatics, Tokyo ♦ Charles University in Prague, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics ◦ Stanford University, Center for the Study of Language and Information sdp-organizers@delph-in.net Abstract We announce a new language resource for research on semantic parsing, a large, carefully curated collection of semantic dependency graphs representing multiple linguistic traditions. This resource is called SDP 2016 and provides an update and extension to previous versions used as Semantic Dependency Parsing target representations in the 2014 and 2015 Semantic Evaluation Exercises (SemEval). For a common core of English text, this third edition comprises semantic dependency graphs from four distinct frameworks, packaged in a unified abstract format and aligned at the sentence and token levels. SDP 2016 is the first general release of this resource and available for licensing from the Linguistic Data Consortium from May 2016. The data is accompanied by an open-source SDP utility toolkit and system results from previous contrastive parsing evaluations against these target representations. Keywords: Semantic Dependency Parsing, Semantic Parsing, Semantic Dependencies, Meaning Representation 1. Background and Goals Increased interest in natural language ‘understanding’ has brought into the focus of much current research a variety of techniques often described as ‘semantic parsing’. In this work, we seek to contribute to improved comparability of representations and results in one sub-area of semantic pars- ing, viz. analysis of natural language strings into bi-lexical semantic dependency graphs; these are general represen- tations of sentence meaning whose nodes correspond to surface lexical units and whose directed edges encode core predicate–argument relations.1 We have prepared for public release in May 2016 an as- sortment of semantic dependency graphs stemming from four distinct linguistic traditions, for substantial volumes of English, Chinese, and Czech running text, and providing both in-domain training and test, as well as out-of-domain test data. To eliminate minor divergences across resources, these graph banks are aligned at the sentence and token level, within each language, represented in a unified abstract graph model, and packaged in a common file format. Where applicable, the graphs are paired with original representa- tions from which they were derived (e.g. underspecified logical-forms or tectogrammatical trees), and with corre- sponding ‘companion’ syntactic analyses from a broad vari- ety of frameworks and sources, comprising gold-standard annotations as well as analyses delivered by state-of-the-art statistical parsers. 1Domain- and application-independence and lexicalization set these target representations apart from other strands of seman- tic parsing, into immediately actionable query languages in the tradition of Zelle & Mooney (1996), on the one hand, or into rep- resentations whose primitives need not be surface lexical units, on the other hand, as for example English Resource Semantics (Copestake & Flickinger, 2000; Flickinger et al., 2014), the Dis- course Representation Structures …

Content truncated to 500 words