IMPROVING SPEED AND PRECISION IN PHRASE BASED INDEXING BY CLAUSAL SEGMENTAION

摘要

Generally the precision is more important in information retrieval (IR) for web documents. For improvement of the precision, we adopt phrase based indexing method using parser, which is used to extract the phrases as indexing unit. But parser encounters some problems. If sentence is long, the parsing process generally requires much time in proportion to sentence length. Also the parsing precision is very low because of much ambiguity in parsing. This paper uses the clausal segmentation technique for solving the problems in our phrase based indexing method. We use dependency rule and context patterns of Korean sentence for clausal segmentation. The experiment result shows that the indexing precision and speed by our method had more improvement than the indexing without segmentation method.