Word vector generation device, word vector generation method, program, and recording medium

摘要

PPROBLEM TO BE SOLVED: To provide a word vector generation device for storing co-occurrence information of all different words with an arbitrary word, suppressing the number of dimensions of the co-occurrence vectors, and generating co-occurrence vectors to reduce the computational complexity of processing to the co-occurrence vectors by suppressing the number of dimensions of the co-occurrence vectors, and to provide a word vector generation method, a program, and a recording medium. PSOLUTION: A word vector generation device is configured to associate some of N components with each of words in a word dictionary, to morphologically analyze an input text, to generate a co-occurrence matrix between the obtained group of different words and the group of N components, and to store a value calculated by adding the frequency that an arbitrary word A co-occurs with arbitrary components B associated with the words in a predetermined range of the text over the predetermined range in the text, into an element in the co-occurrence matrix corresponding to the word A and the components B. PCOPYRIGHT: (C)2010,JPO&INPIT