Ontology-Oriented Information Extraction and Integration

Computational information sources store and provide such large amounts of data that accessing, finding or summarizing information remains a difficult task given the sheer amount of information to be found in each source and given the large number and variety of sources available through current technologies, such as the WWW. The reasons underlying this problem are manifold, however one of the major causes lies in the large gap between the conceptualizations of information, such as envisioned by the user, and the actual storage and provision of information. By and large, the question remains how to bridge this gap at all and how to bridge it in a way that reduces the engineering task of providing information extraction methods for a large number and variety of information sources: free text, semi-structured information (e.g. XML), and relational database information all exhibit similar problems when it comes to providing a common conceptualization for underlying information. Ontologies describe shared conceptualizations for particular domains of interest on a high-level of technical abstraction. Rather then dealing with implementation issues, domain ontologies just describe the concepts relevant to this domain, their relationships, and rules about these relationships that enforce a well-defined semantics on the conceptualization. Regarding the information extraction and integration problem just mentioned, formal ontologies allow the precise description of a conceptualization common to varying information sources. Thus, ontologies offer themselves as (partial) solutions to the problems of extracting information and providing it to the user (or a well-known database) in a clear way (cf. Fig. 1)