Database Analysis for Ontology Learning

This paper proposes a database analysis method aimed at the process of building a new ontology. The analysis is based on the idea that a database already contains indirect knowledge of the domain. By finding information about the values contained in the database tables and fields, it is possible to extract some of this knowledge. The obtained information can then be used as the basis for ontology concept creation. The information consists of statistical information about the values in a field, detected distinct values and their distributions, and implicit foreign-key to primary-key relationship detection. Different aspects of the information about the values in the fields, obtained from the analysis, can be used to create value-based and other ontology concepts. The proposed method has been applied on a medical database, containing records about respondents in a study of gastric cancer risk. The inconsistencies and hurdles from working with a database and ways of dealing with them are also discussed in the paper.