CLIP@UMD at SemEval-2016 Task 8: Parser for Abstract Meaning Representation using Learning to Search

Proceedings of SemEval-2016, pages 1190–1196, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics CLIP@UMD at SemEval-2016 Task 8: Parser for Abstract Meaning Representation using Learning to Search Sudha Rao1,3∗, Yogarshi Vyas1,3∗, Hal Daum´e III1,3, Philip Resnik2,3 1Computer Science, 2Linguistics, 3UMIACS University of Maryland {raosudha,yogarshi,hal}@cs.umd.edu, resnik@umd.edu Abstract In this paper we describe our approach to the Abstract Meaning Representation (AMR) parsing shared task as part of SemEval 2016. We develop a novel technique to parse En- glish sentences into AMR using Learning to Search. We decompose the AMR parsing task into three subtasks - that of predicting the con- cepts, the relations, and the root. Each of these subtasks are treated as a sequence of predic- tions. Using Learning to Search, we add past predictions as features for future predictions, and define a combined loss over the entire AMR structure. 1 Introduction This paper describes our submission to the Abstract Meaning Representation (AMR) Parsing Shared Task at SemEval 2016. The goal of the task is to generate AMRs automatically for English sentences. We develop a novel technique for AMR parsing that uses Learning to Search (L2S) (Ross et al., 2011; Daum´e III et al., 2009; Collins and Roark, 2004). L2S is a family of approaches that solves struc- tured prediction problems. These algorithms have proven to be highly effective for problems in NLP like part-of-speech tagging, named entity recogni- tion (Daum´e III et al., 2014), coreference resolution (Ma et al., 2014), and dependency parsing (He et al., 2013). Briefly, L2S attempts to do structured prediction by (1) decomposing the production of the structured output in terms of an explicit search space (states, actions, etc.); and (2) learning hypotheses ∗The first two authors contributed equally to this work. read-01 i book name forest Stories from Nature ARG0 ARG1 name topic op1 op2 op3 Figure 1: AMR graph for the sentence “I read a book, called Stories from Nature, about the forest.” that control a policy that takes actions in this search space. AMR (Banarescu et al., 2013), in turn, is a structured semantic representation which is a rooted, directed, acyclic graph. The nodes of this graph rep- resent concepts in the given sentence and the edges represent relations between these concepts. As such, the task of predicting AMRs can be naturally placed in the L2S framework. This allows us to model the learning of concepts and relations in a unified set- ting which aims to minimize the loss over the entire predicted structure. In the next section, we briefly review DAGGER and explain its various components with respect to our AMR parsing task. Section 3 describes our main algorithm along with the strategies we use to deal with the large search space of the search problem. 1190 I read a book called Stories from Nature i read-01 NULL book NULL call-01 called Stories story NULL from NULL Nature NULL (a) Concept prediction stage: Shaded nodes indicate predicted concepts (Current state). The middle row represents the oracle action. Other rows …

Content truncated to 500 words