Kazunori Matsumoto


pdf bib
Learning Cooccurrences by using a Parser
Kazunori Matsumoto | Hiroshi Sakaki | Shingo Kuroiwa
Proceedings of the First International Workshop on Parsing Technologies

This paper describes two methods for the acquisition and utilization of lexical cooccurrence relationships. Under these method, cooccurrence relationships are obtained from two kinds of inputs: example sentences and the corresponding correct syntactic structure. The first of the two methods treats a set of governors each element of which is bound to a element of sister nodes set in a syntactic structure under consideration, as a cooccurrence relationship. In the second method, a cooccurrence relationship name and affiliated attribute names are manually given in the description of augmented rewriting rules. Both methods discriminate correctness of cooccurrence by the use of the correct syntactic structure mentioned above. Experiment is made for both methods to find if thus obtained cooccurrence relationship is useful for the correct analysis.