Grammar fragment acquisition using syntactic and semantic clustering

A new method for automatically acquiring grammar fragments for understanding uently spoken language is proposed. The goal of this method is to generate a collection of grammar fragments each representing a set of syntactically and semantically similar phrases. First phrases observed frequently in the training set are selected as candidates. Each candidate phrase has three associated probability distributions: of succeeding contexts, of preceding contexts, and of associated machine actions. The similarity between candidate phrases is measured by applying the Kullback-Leibler distance to three probability distributions. Candidate phrases which are close in all three distances are clustered into a grammar fragment. This approach detected 246 phrases in the test-set that were not present in the training-set. Experimental results show that a 3% improvement in the call-type classi cation performance has been achieved by introducing these fragments. key words spoken understanding, preceding and succeeding contexts, Kullback-Leibler distance, phrase similarity, phrase clustering