SudocAD: A Knowledge-Based System for Object Identification

Object identification problems aim to distinguish, to recognize in a computer system symbols aiming at representing the same individual entity in the world modeled in the system. These problems acquired a new importance due to the web, especially in the so-called web of objects, where, for instance, one wants to gather all available information concerning the same entity. Most solutions of object identification problems are based on classification techniques. Recently, logical approaches using knowledge representation techniques have been introduced and combinations of these two kinds of methods have been developed. In this paper we present such a mixed method for the semi-automatic identification of an author in a authority catalog from a bibliographic metadata. The proposed method has been impelemented and evaluated on data extracted from french libraries.