Semantic Interpretation of Compound Nominalization Using TreeBank and the World Wide Web

The interpretation of nominal compounds is one of the most difficult problems in natural language processing. This paper proposes a new model for the automatic classification of four coarse-grained semantic relations involved in Chinese compound nominalizations. In such a model, for a compound nominalization, its paraphrased syntactic role occurrences (PSRO) in a treebank are exploited to form feature vectors for supervised classifiers. To solve the problem of data sparseness, the World Wide Web is used to discover relational clusters and such clusters are employed to produce smoothed PSRO feature vectors for the compound nominalizations. The experimental results show that such a method is very effective.