CLASSIFICATION OF INFORMATION TOPICS BY CLUSTERING INTEREST PROFILES

摘要

A computer program was applied to cluster indexing terms into field-of-interest categoris, defined responses of staff members of a personnel research laboratory. This provides a practical scheme for document classification. The method clusters successively pairs of topics that have the highest probability of being marked together by the scientist as both being of interest or neither one of interest. Ten fields of interest related to the group mission were identified by this hierarchal grouping.