METHOD AND APPARATUS FOR SELECTING IMPORTANT FEATURE OF MULTI-LABEL DATA APPARATUS FOR SELECTING IMPORTANT WORD OF MULTI-CATEGORY DOCUMENT

摘要

The present invention relates to a method and a device to sort key characteristics of multilabel data, and a device to sort key words of a multicategory document. The method includes: a step in which an importance score calculating part calculates an importance score for each characteristic based on a mutual information quantity between all key labels selected from every label group in accordance with preset standards and each of all characteristics about multilabel data; a step in which an approximate importance score calculating part calculates an approximate importance score for each of the characteristics based on an approximate mutual information quantity between all nonimportant labels selected in accordance with preset standards and each of all the characteristics; and a step in which a total importance score calculating part calculates a total importance score for each of the characteristics by adding up the importance and approximate importance scores for each of all the characteristics.