Prediction of Maximal Projection for Semantic Role Labeling

Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), pages 833–840 Manchester, August 2008 Prediction of Maximal Projection for Semantic Role Labeling Weiwei Sun∗, Zhifang Sui Institute of Computational Linguistics Peking University Beijing, 100871, China {ws, szf}@pku.edu.cn Haifeng Wang Toshiba (China) R&D Center 501, Tower W2, Oriental Plaza Beijing, 100738, China wanghaifeng@rdc.toshiba.com.cn Abstract In Semantic Role Labeling (SRL), argu- ments are usually limited in a syntax sub- tree. It is reasonable to label arguments lo- cally in such a sub-tree rather than a whole tree. To identify active region of argu- ments, this paper models Maximal Pro- jection (MP), which is a concept in D- structure from the projection principle of the Principle and Parameters theory. This paper makes a new definition of MP in S- structure and proposes two methods to pre- dict it: the anchor group approach and the single anchor approach. The anchor group approach achieves an accuracy of 87.75% and the single anchor approach achieves 83.63%. Experimental results also indicate that the prediction of MP improves seman- tic role labeling. 1 Introduction Semantic Role Labeling (SRL) has gained the in- terest of many researchers in the last few years. SRL consists of recognizing arguments involved by predicates of a given sentence and labeling their semantic types. As a well defined task of shallow semantic parsing, SRL has a variety of applications in many kinds of NLP tasks. A variety of approaches has been proposed for the different characteristics of SRL. More re- cent approaches have involved calibrating features (Gildea and Jurafsky, 2002; Xue and Palmer, 2004; ∗This work was partial completed while this author was at Toshiba (China) R&D Center. ∗c⃝ 2008. Licensed under the Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported li- cense (http://creativecommons.org/licenses/by-nc-sa/3.0/). Some rights reserved. Pradhan et al., 2005), analyzing the complex input – syntax trees (Moschitti, 2004; Liu and Sarkar, 2007), exploiting the complicated output – the predicate-structure (Toutanova et al., 2005), as well as capturing paradigmatic relations between predicates (Gordon and Swanson, 2007). In prior SRL methods, role candidates are ex- tracted from a whole syntax tree. Though sev- eral pruning algorithms have been raised (Xue and Palmer, 2004), the policies are all in global style. In this paper, a statistical analysis of Penn Prop- Bank indicates that arguments are limited in a local syntax sub-tree rather than a whole one. Prior SRL methods do not take such locality into account and seek roles in a wider area. The neglect of local- ity of arguments may cause labeling errors such as constituents outside active region of arguments may be falsely recognized as roles. This paper uses insights from generative lin- guistics to guide the solution of locality of argu- ments. In particular, Maximal Projection (MP) which dominates1 active region of arguments ac- cording to the projection principle of principle and parameters. Two methods, the anchor group ap- proach and the single anchor approach, are pro- posed to find the active sub-tree which is rooted by MP and covers all roles. …

Content truncated to 500 words