Support Vector Machine Applied to the Semantic Interpretation of VN Compound

The semantic interpretation of nominal compounds is one of the most difficult problems in natural language processing. VN compound is a subset of nominal compounds where the modifier is a verb nominalization. This paper proposes a new interpretation model in which a support vector machine is applied to label five semantic relations involved in Chinese VN compounds. The World Wide Web is exploited as a large corpus to compute point-wise mutual information between the VN compounds and a set of relation specific lexical patterns. Such Web-based statistics is used as the classification features for the support vector machine. By applying a sub-linear transformation and discretization of the raw statistics, a good result is obtained for the five semantic relations.