Automated Subject Analysis of Short Phrase by the Method of Comparison with the Bibliographic Records in E-Catalog

Abstract

There is proposed the new method of automated subject classification in a full-sized natural rubricator, that can be applied to ultra short texts. The method is based on the use of the adopted classification decisions, borrowed from bibliographic records. The solutions found are specified by the information received from the classification system, and generalized by different ways. There are considered four options for obtaining classification decision, which differ by the number of failures and the proportion of incorrect results. There is demonstrated successful method for subject analysis of titles of works by LBC divisions. There are considered the problems of bibliographic catalogue and classification system, reducing the result. There are presented possible negative consequences of using automated classification for the quality of bibliographic catalog.