Word weighting method and apparatus, word weighting program, and computer-readable recording medium recording word weighting program

摘要

PPROBLEM TO BE SOLVED: To enable weighting which reflects semantic of word rather than a method on the basis of only notation of word. PSOLUTION: In this method, text is divided into a word unit; distribution is taken which is 1 at an appearance position of a word w and 0 at other positions for every different word w; a probability density function is estimated which uses the distribution as an observed value distribution; and the weight of word w is computed on the basis of entropy of the probability density function. PCOPYRIGHT: (C)2005,JPO&NCIPI