End-to-end learning of semantic role labeling using recurrent neural networks

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 1127–1137, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics End-to-end Learning of Semantic Role Labeling Using Recurrent Neural Networks Jie Zhou and Wei Xu Baidu Research {zhoujie01,wei.xu}@baidu.com Abstract Semantic role labeling (SRL) is one of the basic natural language processing (NLP) problems. To this date, most of the suc- cessful SRL systems were built on top of some form of parsing results (Koomen et al., 2005; Palmer et al., 2010; Pradhan et al., 2013), where pre-defined feature tem- plates over the syntactic structure are used. The attempts of building an end-to-end SRL learning system without using pars- ing were less successful (Collobert et al., 2011). In this work, we propose to use deep bi-directional recurrent network as an end-to-end system for SRL. We take on- ly original text information as input fea- ture, without using any syntactic knowl- edge. The proposed algorithm for seman- tic role labeling was mainly evaluated on CoNLL-2005 shared task and achieved F1 score of 81.07. This result outperforms the previous state-of-the-art system from the combination of different parsing trees or models. We also obtained the same conclusion with F1 = 81.27 on CoNLL- 2012 shared task. As a result of simplicity, our model is also computationally efficient that the parsing speed is 6.7k tokens per second. Our analysis shows that our model is better at handling longer sentences than traditional models. And the latent vari- ables of our model implicitly capture the syntactic structure of a sentence. 1 Introduction Semantic role labeling (SRL) is a form of shal- low semantic parsing whose goal is to discover the predicate-argument structure of each predicate in a given input sentence. Given a sentence, for each target verb (predicate) all the constituents in the sentence which fill a semantic role of the verb have to be recognized. Typical semantic argu- ments include Agent, Patient, Instrument, etc., and also adjuncts such as Locative, Temporal, Man- ner, Cause, etc.. SRL is useful as an intermedi- ate step in a wide range of natural language pro- cessing (NLP) tasks, such as information extrac- tion (Bastianelli et al., 2013), automatic document categorization (Persson et al., 2009) and question- answering (Dan and Lapata, 2007; Surdeanu et al., 2003; Moschitti et al., 2003). SRL is considered as a supervised machine learning problem. In traditional methods, linear classifier such as SVM is often employed to per- form this task based on features extracted from the training corpus. Actually, people often treat this problem as a multi-step classification task. First, whether an argument is related to the predicate is determined; next the detail relation type was de- cided(Palmer et al., 2010). Syntactic information is considered to play an essential role in solving this problem (Punyakanok et al., 2008a). The location of an argument on syn- tactic tree provides an intermediate tag for improv- ing the performance. However, building this syn- tactic tree …

Content truncated to 500 words