Capturing Argument Relationship for Chinese Semantic Role Labeling

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2011–2016, Austin, Texas, November 1-5, 2016. c⃝2016 Association for Computational Linguistics Capturing Argument Relationships for Chinese Semantic Role Labeling Lei Sha, Tingsong Jiang, Sujian Li, Baobao Chang, Zhifang Sui Key Laboratory of Computational Linguistics, Ministry of Education School of Electronics Engineering and Computer Science, Peking University Collaborative Innovation Center for Language Ability, Xuzhou 221009 China shalei, tingsong, lisujian, chbb, szf@pku.edu.cn Abstract In this paper, we capture the argument rela- tionships for Chinese semantic role labeling task, and improve the task’s performance with the help of argument relationships. We split the relationship between two candidate argu- ments into two categories: (1) Compatible ar- guments: if one candidate argument belongs to a given predicate, then the other is more likely to belong to the same predicate; (2) In- compatible arguments: if one candidate argu- ment belongs to a given predicate, then the other is less likely to belong to the same predi- cate. However, previous works did not explic- itly model argument relationships. We use a simple maximum entropy classifier to capture the two categories of argument relationships and test its performance on the Chinese Propo- sition Bank (CPB). The experiments show that argument relationships is effective in Chinese semantic role labeling task. 1 Introduction Semantic Role Labeling (SRL) is defined as the task to recognize arguments for a given predicate and as- sign semantic role labels to them. Because of it- s ability to encode semantic information, there has been an increasing interest in SRL on many lan- guages (Gildea and Jurafsky, 2002; Sun and Juraf- sky, 2004). Figure 1 shows an example in Chinese Proposition Bank (CPB) (Xue and Palmer, 2003), which is a Chinese corpus annotated with semantic role labels. Previous works of Chinese SRL include feature- based approaches and neural network based ap- proaches. Feature-based approaches often extract a large number of handcrafted features from the sen- tence, and feed these features to statistical classifiers such as CRF, MaxEnt and SVM (Sun and Jurafsky, 2004; Xue, 2008; Ding and Chang, 2008; Ding and Chang, 2009; Sun, 2010). Neural network based approaches usually take Chinese SRL as sequence labeling task and use bidirectional recurrent neural network (RNN) with long-short-term memory (LST- M) to solve the problem (Wang et al., 2015). However, both of the above two kinds of ap- proaches identify each candidate argument separate- ly without considering the relationship between ar- guments. We define two categories of argument re- lationships here: (1) Compatible arguments: if one candidate argument belongs to a given predicate, then the other is more likely to belong to the same predicate; (2) Incompatible arguments: if one can- didate argument belongs to a given predicate, then the other is less likely to belong to the same pred- icate. For example, in Figure 1, the word “� û”(foreign businessman) and “è’”(entrepreneur) tend to be compatible arguments when the predi- cate word is “Ý]”(invest). On the other hand, “è ’”(entrepreneur) and “ 5½”(rule) are …

Content truncated to 500 words