Engineering Semantic Web Ontologies by Ontology Merging and Classification
The Semantic Web uses ontology as a key technology for achieving interoperability across heterogeneous applications and enabling better environment for human and machines to communicate and work collaboratively. Ontologies on the semantic web standardize and formalize the meaning of terminology through concepts and store domain knowledge in a machine understandable and processable manner. But, the number of ontologies being developed and maintained over the web is significantly increasing, which demands various new techniques for the ontology storage, classification, merging, reusability, searching, etc. Therefore, for the realization of Semantic Web vision, there have been a lot of more efforts needed to fulfill its promises. This thesis explores two tasks for managing multiple ontologies on the semantic web, i.e., merging of heterogeneous ontologies and their classification. Recent years have witnessed great development of ontology merging techniques and systems. They act as pillars in the wide range of application domains and building collaborations that involve the sharing of data, knowledge and resources among modern companies, and aid in developing a new ontology by reusing existing open ontologies. Although there is a great effort seen, still the state-of-the-art on ontology merging systems is semi-automatic. It reduces the burden of manual creation and maintenance of mappings and needs human intervention for their validation and merging. This thesis presents a semantic merge framework that exploits novel automatic methodologies for the detection of semantic inconsistencies and conflict resolution strategies in the ontology merging process and thus minimizes human involvement one more step down during the ontology merging process. Its ultimate goal is to check the semantic correctness and consistency of mappings, and to ensure the satisfiability of the merged ontology. The proposed framework is implemented and evaluated on various real world test cases with encouraging results, thus, proving empirically its benefits. As our framework exploits all types of semantics provided in the source ontologies and employs a test criteria for the validation of initial mappings found in the initial stages, our approach enhances the accuracy of mapping and merging ontologies, and produces a consistent, complete and coherent global ontology from the local heterogeneous ontologies. In this way, it forms a global layer from which several heterogeneous local ontologies could be accessed and hence would exchange information in semantically sound manners. The proposed merging methodology is applied on virtual warehouse, where data resides physically in the original sources. Due to the intermediate mediation layer that tackles the semantic heterogeneities transparently to meet the user request, often the generated results of data integration suffer from inconsistency, incompleteness and redundancy. Therefore, in our case study, we focus on ontology based query answering based on a reliable mediator that builds Quality Criterion for the query answering. We learn that the quality of a global ontology has as strong connection with the query results obtained because when the global ontology is erroneous having inconsistencies, incompleteness or redundancies then query answering would heavily compromise results. Our case study improves query answering based on semantics from underlying ontologies and provided mechanisms to find more implicit information about data sources. Semantic mapping between concepts allows inferred extraction of other types of implicit information from semantic paths between data sources. User queries are transformed into local queries which can provide more meaningful results and a better meeting with the intention of the user. We conclude that quality criteria based on consistency, completeness and conciseness is indeed a suitable model for supporting the reliable virtual warehouse. This thesis explores another vital task about the classification of web ontologies. Now-a-days, the semantic web has gained huge momentum with the explosive number of ontologies, where multiple ontologies associated with a same domain/concept appear to be quite common. Therefore, it is of immense importance to classify the ontologies into respective domain hierarchies. It helps humans and web agents to find the correct and desired ontology (or concept) on the web and supports ontology engineering processes. Ontology Classification is also essential for many other tasks such as development of ontology directories on the web, focused crawling for ontology retrieval, concept specific modular ontology analysis, improving quality of search, etc. In order to meet the real challenge of ontology searching and retrieval, this thesis presents an ontology-based approach for the classification of web ontologies. We believe that once ontologies are classified properly, then they are searched in a sound semantic manner on the Semantic Web. For building a classification approach, we benefit from our approach of ontology matching with several modifications. We believe that an ontology-based approach works better for overlapping ontologies that come across due to the semantic heterogeneity and knowledge structure requirement during the domain modeling. Ontology classification is necessary for the construction, maintenance or expansion of ontology directories on the semantic web. Currently, ontology directories are maintained by human editors that facilitate users to browse for ontologies within a predefined set of categories. Ontology classifier does this job automatically replacing tedious manual effort to help update and expand such directories.
KEYWORDS: Ontology Merging; Ontology Classification; Interoperability of Information Systems; Heterogeneous Systems