UTDMet: Combining WordNet and Corpus Data for Argument Coercion Detection

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 252–255, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics UTDMet: Combining WordNet and Corpus Data for Argument Coercion Detection Kirk Roberts and Sanda Harabagiu Human Language Technology Research Institute University of Texas at Dallas Richardson, Texas, USA {kirk,sanda}@hlt.utdallas.edu Abstract This paper describes our system for the classification of argument coercion for SemEval-2010 Task 7. We present two ap- proaches to classifying an argument’s se- mantic class, which is then compared to the predicate’s expected semantic class to detect coercions. The first approach is based on learning the members of an arbi- trary semantic class using WordNet’s hy- pernymy structure. The second approach leverages automatically extracted seman- tic parse information from a large corpus to identify similar arguments by the pred- icates that select them. We show the re- sults these approaches obtain on the task as well as how they can improve a tradi- tional feature-based approach. 1 Introduction Argument coercion (a type of metonymy) occurs when the expected semantic class (relative to the a predicate) is substituted for an object of a dif- ferent semantic class. Metonymy is a pervasive phenomenon in language and the interpretation of metonymic expressions can impact tasks from se- mantic parsing (Scheffczyk et al., 2006) to ques- tion answering (Harabagiu et al., 2005). A seminal example in metonymy from (Lakoff and Johnson, 1980) is: (1) The ham sandwich is waiting for his check. The ARG1 for the predicate wait is typically an animate, but the “ham sandwich” is clearly not an animate. Rather, the argument is coerced to ful- fill the predicate’s typing requirement. This coer- cion is allowed because an object that would nor- mally fulfill the typing requirement (the customer) can be uniquely identified by an attribute (the ham sandwich he ordered). SemEval-2010 Task 7 (“Argument Selection and Coercion”) (Pustejovsky and Rumshisky, 2009) was designed to evaluate systems that de- tect such coercions and provide a “compositional history” of argument selection relative to the pred- icate. In order to accomplish this, an argument is annotated with both the semantic class to which it belongs (the “source” type) as well as the class ex- pected by the predicate (the “target” type). How- ever, in the data provided, the target type was un- ambiguous given the lemmatized predicate, so the remainder of this paper discusses source type clas- sification. The detection of coercion is then sim- ply performed by checking if the classified source type and target type are different. In our system, we explore two approaches with separate underlying assumptions about how arbi- trary semantic classes can be learned. In our first approach, we assume a semantic class can be de- fined a priori from a set of seed terms and that WordNet is capable of defining the membership of that semantic class. We apply the PageRank al- gorithm in order to weight WordNet synsets given a set of seed concepts. In our second approach, we assume that arguments …

Content truncated to 500 words