How to recognize interesting topics to provide cooperative answering

Abstract

We present a method, and its formalization, to provide interesting additional information to users asking queries to a relational data base. The interesting information is defined using a knowledge base containing rules which represent the expertise of an expert having a long experience in providing information to casual users. This knowledge base also contains a high-level description of the data base content. This description uses the concepts of entity, attribute of entity, relationship and topic. The topics are related to attributes and relationships in order to be able to characterize all the information in the data base which belongs to a given semantic field. The data base and knowledge base are both formalized in first-order logic. There are two different formalization levels: the Object level to represent the data base itself, and the Meta level to represent the knowledge base which is used to transform queries. The transformed queries define the additional information which is provided to users. They can be obtained with a standard deduction mechanism. A first prototype implemented in Prolog is running.