An Approach to Classify Semi-structured Objects

Several advanced applications, such as those dealing with the Web, need to handle data whose structure is not known a-priori. Such requirement severely limits the applicability of traditional database techniques, that are based on the fact that the structure of data (e.g. the database schema) is known before data are entered into the database. Moreover, in traditional database systems, whenever a data item (e.g. a tuple, an object, and so on) is entered, the application specifies the collection (e.g. relation, class, and so on) the data item belongs to. Collections are the basis for handling queries and indexing and therefore a proper classification of data items in collections is crucial. In this paper, we address this issue in the context of an extended object-oriented data model. We propose an approach to classify objects, created without specifying the class they belong to, in the most appropriate class of the schema, that is, the class closest to the object state. In particular, we introduce the notion of weak membership of an object in a class, and define two measures, the conformity and the heterogeneity degrees, exploited by our classification algorithm to identify the most appropriate class in which an object can be classified, among the ones of which it is a weak member.