Co-occurrence expression extraction apparatus and co-occurrence expression extraction method

摘要

PPROBLEM TO BE SOLVED: To solve a problem of a conventional co-occurrence expression extraction based on a frequency, where the great number of redundant co-occurrence expressions between general words are extracted while, in a difference analysis as a method for eliminating redundant combinations, a layer from which a difference is extracted is not self-evident when prior knowledge such as the type of opinion cannot be used. PSOLUTION: In a co-occurrence expression extracting device, a feature section is given by using a temporal change as a reference for a candidate word extracted from text data having time series information such as writing in a web bulletin board, and when difference analysis is carried out between the text of the feature section and that of a non-feature section, co-occurrence extraction with greatly reduced redundant combinations can be achieved. PCOPYRIGHT: (C)2011,JPO&INPIT