Priberam: A Turbo Semantic Parser with Second Order Features

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 471–476, Dublin, Ireland, August 23-24, 2014. Priberam: A Turbo Semantic Parser with Second Order Features Andr´e F. T. Martins∗† Mariana S. C. Almeida∗† ∗Priberam Labs, Alameda D. Afonso Henriques, 41, 2o, 1000-123 Lisboa, Portugal †Instituto de Telecomunicac¸˜oes, Instituto Superior T´ecnico, 1049-001 Lisboa, Portugal {atm,mla}@priberam.pt Abstract This paper presents our contribution to the SemEval-2014 shared task on Broad- Coverage Semantic Dependency Parsing. We employ a feature-rich linear model, in- cluding scores for first and second-order dependencies (arcs, siblings, grandparents and co-parents). Decoding is performed in a global manner by solving a linear relax- ation with alternating directions dual de- composition (AD3). Our system achieved the top score in the open challenge, and the second highest score in the closed track. 1 Introduction The last decade saw a considerable progress in sta- tistical modeling for dependency syntactic pars- ing (K¨ubler et al., 2009). Models that incorporate rich global features are typically more accurate, even if pruning is necessary or decoding needs to be approximate (McDonald et al., 2006; Koo and Collins, 2010; Bohnet and Nivre, 2012; Martins et al., 2009, 2013). This paper applies the same ratio- nale to semantic dependency parsing, in which the output variable is a semantic graph, rather than a syntactic tree. We extend a recently pro- posed dependency parser, TurboParser (Martins et al., 2010, 2013), to be able to perform semantic parsing using any of the three formalisms consid- ered in this shared task (DM, PAS, and PCEDT). The result is TurboSemanticParser, which we re- lease as open-source software.1 We describe here a second order model for se- mantic parsing (§2). We follow prior work in se- mantic role labeling (Toutanova et al., 2005; Jo- This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and pro- ceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ 1http://labs.priberam.com/Resources/ TurboSemanticParser Figure 1: Example of a semantic graph in the DM formalism (sentence #22006003). We treat top nodes as a special semantic role TOP whose predi- cate is a dummy root symbol. hansson and Nugues, 2008; Das et al., 2012; Flani- gan et al., 2014), by adding constraints and model- ing interactions among arguments within the same frame; however, we go beyond such sibling in- teractions to consider more complex grandpar- ent and co-parent structures, effectively correlat- ing different predicates. We formulate parsing as a global optimization problem and solve a relax- ation through AD3, a fast dual decomposition al- gorithm in which several simple local subprob- lems are solved iteratively (§3). Through a rich set of features (§4), we arrive at top accuracies at parsing speeds around 1,000 tokens per second, as described in the experimental section (§5). 2 A Second Order Model for Parsing Figure 1 depicts a sentence and its semantic graph. We cast semantic parsing as a structured predic- tion problem. Let x be a sentence and Y(x) the set of possible dependency graphs. We assume each candidate …

Content truncated to 500 words