CU-NLP at SemEval-2016 Task 8: AMR Parsing using LSTM-based Recurrent Neural Networks

Proceedings of SemEval-2016, pages 1197–1201, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics CU-NLP at SemEval-2016 Task 8: AMR Parsing using LSTM-based Recurrent Neural Networks William R. Foland Jr. OK Robot Go, Ltd. 5345 Dunraven Circle Golden, Co, 80403, USA bill.foland@okrobotgo.com James H. Martin Department of Computer Science and Institute of Cognitive Science University of Colorado Boulder, CO 80309 James.Martin@colorado.edu Abstract We describe the system used in our participa- tion in the AMR Parsing task for SemEval- 2016. Our parser does not rely on a syntactic pre-parse, or heavily engineered features, and uses five recurrent neural networks as the key architectural components for estimating AMR graph structure. 1 Introduction Abstract Meaning Representation, or AMR (Ba- narescu et al., 2012) is a graph-based representation of the meaning of sentences which incorporates lin- guistic phenomena such as semantic roles, corefer- ence, negation, and more.1 The process of creating AMR’s for sentences is called AMR Parsing. We used an early version of the system described in this paper to generate our submission to the Semeval-2016 Meaning Represen- tation Parsing Task.2 The details of our system will be explained using this example sentence: France plans further nu- clear cooperation with numerous countries . A graphical depiction is shown in Figure 1. The system extracts features from the sentence which are processed by a form of recurrent neural network called BDLSTM to create a set of AMR concepts. Features from these concepts are pro- cessed by a pair of BDLSTM networks to compute relation probabilities. All concepts are then con- nected using an iterative, greedy algorithm to com- pute the set of relations in the AMR. Another two 1http://amr.isi.edu/language.html 2http://alt.qcri.org/semeval2016/task8/# BDLSTM networks compute attribute and name cat- egories to complete the estimation of AMR element probabilities. 2 Related Work Most current AMR parsers assume input that has undergone varying degrees of syntactic analysis, ranging from simple part-of-speech tagging to more complex dependency or phrase-structure analysis. (Wang et al., 2015; Vanderwende et al., 2015; Peng et al., 2015; Pust et al., 2015; Artzi et al., 2015; Flanigan et al., 2014; Werling et al., 2015). In con- trast, we follow the spirit of minimal feature extrac- tion using pre-trained word embeddings, as in (Col- lobert et al., 2011) and a recurrent network archi- tecture similar to that described in (Zhou and Xu, 2015). 3 System Architecture 3.1 Feature Extraction In our system, all features are represented by em- bedding vectors, trained and stored in lookup ta- bles. Word feature embeddings are mapped from the words in the sentence, and are trained with back propagation just like other parameters in the net- work. They are initialized with vectors which are pre-trained on large corpora of english text, we use the word embeddings from (Collobert et al., 2011). The only explicit features not derived from the raw input are features based on named entity recog- nition (NER). We first use the Univ. of Illinois Wik- ifier to find and classify named entities and then en- code …

Content truncated to 500 words