tailieunhanh - Báo cáo khoa học: "Novel Semantic Features for Verb Sense Disambiguation Dmitriy Dligach"

We propose a novel method for extracting semantic information about a verb's arguments and apply it to Verb Sense Disambiguation (VSD). We contrast this method with two popular approaches to retrieving this informa tion and show that it improves the performance of our VSD system and outperforms the other two approaches | Novel Semantic Features for Verb Sense Disambiguation Dmitriy Dligach The Center for Computational Language and Education Research 1777 Exposition Drive Boulder Colorado 80301 @ Martha Palmer Department of Linguistics University of Colorado at Boulder 295 UCB Boulder Colorado 80309 @ Abstract We propose a novel method for extracting semantic information about a verb s arguments and apply it to Verb Sense Disambiguation VSD . We contrast this method with two popular approaches to retrieving this information and show that it improves the performance of our VSD system and outperforms the other two approaches 1 Introduction The task of Verb Sense Disambiguation VSD consists in automatically assigning a sense to a verb target verb given its context. In a supervised setting a VSD system is usually trained on a set of pre-labeled examples the goal of this system is to tag unseen examples with a sense from some sense inventory. An automatic VSD system usually has at its disposal a diverse set of features among which the semantic features play an important role verb sense distinctions often depend on the distinctions in the semantics of the target verb s arguments Hanks 1996 . Therefore some method of capturing the semantic knowledge about the verb s arguments is crucial to the success of a VSD system. The approaches to obtaining this kind of knowledge can be based on extracting it from ele c-tronic dictionaries such as WordNet Fellbaum 1998 using Named Entity NE tags or a combi nation of both Chen 2005 . In this paper we propose a novel method for obtaining semantic knowledge about words and show how it can be applied to VSD. We contrast this method with the other two approaches and compare their performances in a series of experiments. 2 Lexical and Syntactic Features We view VSD as a supervised learning problem solving which requires three groups of features lexical syntactic and semantic. Lexical features include all