Semantic role labeling with Boosting, SVMs, Maximum Entropy, SNOW, and Decision Lists

Semantic Role Labeling with Boosting, SVMs, Maximum Entropy, SNOW, and Decision Lists Grace NGAI†1, Dekai WU‡2 Marine CARPUAT‡, Chi-Shing WANG†, Chi-Yung WANG† † Dept. of Computing HK Polytechnic University Hong Kong ‡ HKUST, Dept of Computer Science Human Language Technology Center Hong Kong csgngai@polyu.edu.hk, dekai@cs.ust.hk marine@cs.ust.hk, wcsshing@netvigator.com, cscywang@comp.polyu.edu.hk Abstract This paper describes the HKPolyU-HKUST sys- tems which were entered into the Semantic Role La- beling task in Senseval-3. Results show that these systems, which are based upon common machine learning algorithms, all manage to achieve good performances on the non-restricted Semantic Role Labeling task. 1 Introduction This paper describes the HKPolyU-HKUST sys- tems which participated in the Senseval-3 Semantic Role Labeling task. The systems represent a diverse array of machine learning algorithms, from decision lists to SVMs to Winnow-type networks. Semantic Role Labeling (SRL) is a task that has recently received a lot of attention in the NLP community. The SRL task in Senseval-3 used the Framenet (Baker et al., 1998) corpus: given a sentence instance from the corpus, a system’s job would be to identify the phrase constituents and their corresponding role. The Senseval-3 task was divided into restricted and non-restricted subtasks. In the non-restricted subtask, any and all of the gold standard annotations contained in the FrameNet corpus could be used. Since this includes information on the boundaries of the parse constituents which correspond to some frame element, this effectively maps the SRL task to that of a role-labeling classification task: given a constituent parse, identify the frame element that it belongs to. Due to the lack of time and resources, we chose to participate only in the non-restricted subtask. This enabled our systems to take the classification ap- proach mentioned in the previous paragraph. 1The author would like to thank the Hong Kong Polytechnic University for supporting this research in part through research grants A-PE37 and 4-Z03S. 2The author would like to thank the Hong Kong Research Grants Council (RGC) for supporting this research in part through research grants RGC6083/99E, RGC6256/00E, and DAG03/04.EG09. 2 Experimental Features This section describes the features that were used for the SRL task. Since the non-restricted SRL task is essentially a classification task, each parse con- stituent that was known to correspond to a frame element was considered to be a sample. The features that we used for each sample have been previously shown to be helpful for the SRL task (Gildea and Jurafsky, 2002). Some of these features can be obtained directly from the Framenet annotations: • The name of the frame. • The lexical unit of the sentence — i.e. the lex- ical identity of the target word in the sentence. • The general part-of-speech tag of the target word. • The “phrase type” of the constituent — i.e. the syntactic category (e.g. NP, VP) that the con- stituent falls into. • The “grammatical function” (e.g. subject, ob- ject, modifier, etc) of the constituent, with re- spect to the target word. • The position (e.g. before, after) of the con- …

Content truncated to 500 words