Collaborative ontology-based information indexing and retrieval

Abstract

An information system like the Web is a continuously evolving system consisting of multiple heterogeneous information sources covering a wide domain of discourse and a huge number of users human or software with diverse characteristics and needs that produce and consume information. The challenge nowadays is to build a scalable information infrastructure enabling the effective, accurate, content-based retrieval of information, in a way that adapts to the characteristics and interests of the users. The aim of this work is to propose formally sound methods for building such an information network based on ontologies which are widely used and are easy to grasp by ordinary Web users. The main results of this work are: - A novel scheme for indexing and retrieving objects according to multiple aspects or facets. The proposed scheme is a faceted scheme enriched with a method for specifying the combinations of terms that are valid. We give a model-theoretic interpretation to this model and we provide mechanisms for inferring the valid combinations of terms. This inference service can be exploited for preventing errors during the indexing process, which is very important especially in the case where the indexing is done collaboratively by many users, and for deriving "complete" navigation trees suitable for browsing through the Web. The proposed schem has several advantages over the hierarchical classification schemes currently employed by Web catalogs, namely, conceptual clarity (it is easier to understand), compactness (it takes less space), and scalability (the update operations can be formulated more easily and be performed more efficently. An exible and efficient model for buliding mediators over ontology based information sources. The proposed mediators support several modes of query translation and evaluation which can accommodate various application needs and levels of answer quality. The proposed model can be used for providing users with customized views of Web catalogs. It can also complement the techniques for building mediators over relational sources as to support approximate translation of partially ordered domain values. - A data driven method for articulating ontologies. This method can be used for the automatic, or semi-automatic, construction of an articulation between two or more materialized ontologies. A distinctive feature of this method is that it is independent of the nature of the objects, i.e. the objects may be images, audio, video, etc, and that it can be implemented efficiently by a communication protocol, thus the involved ontologies (sources) can be distant. - A novel method for fusing the results of sources which return ordered sets of objects. The proposed method is based solely on the actual results which are returned by each source for every query. The final (fused) ordering of objects is derived by aggregating the orderings of each source by a voting process. In addition, the fused ordering is accompanied by a level of agreement (alternatively construed as the level of confidence). The proposed method does not require any prior knowledge about the underlying sources, therefore it is appropriate for environments where the underlying sources are heterogeneous and autonomous, thus it is appropriate for the Web, for example, for building mediators over search engines. These results allow creating a complex network of sources consisting of primary and secondary sources, where a primary source can be ontology-based, retrieval or hybrid (i.e. a source that is both ontology-based and retrieval).