Semantically-enhanced information retrieval model based on multiple merged and dynamically enriched ontologies

Ontologies are used in various Information Retrieval (IR) and Artificial Intelligence (AI) domains and applications such as the Semantic Web (SW), Question Answering (QA), knowledge representation and management, Query Expansion (QE), Natural Language Processing (NLP), and so on. They aim at providing a commonly agreed upon understanding of several domains across different communities. In addition, they define concepts and constraints on their use within a specific domain in a formal and explicit manner. Hence, they are considered as the key element in enabling interoperability between heterogeneous systems and across various applications. However, the decentralized process of ontology development and the differences in viewpoints between ontology engineers have resulted in the so called the "semantic heterogeneity" problem between ontologies. In this context, conflicts in the semantic relations as well as other mismatches can be found between the concepts of ontologies that are developed to encode knowledge about the same domain. For example, we may find two or more domain-specific ontologies that use different terms to refer to the same concept or use the same term to refer to different concepts. To overcome “semantic heterogeneity” and achieve interoperability between heterogeneous systems, we need to resolve the semantic conflicts and other semantic mismatches between similar and overlapping ontologies. Another key challenge that needs to be addressed is how to maintain and dynamically enrich the merged ontologies and keep them up-to-date. To do this, we present a dynamic ontology enrichment model, which integrates semantic and statistical based relatedness measures to enrich ontologies with semantically related concepts and instances. Additionally, this thesis explains how ontological background knowledge (represented by multiple merged and further enriched ontologies from various domains) can be reused to support semantic search and retrieval capabilities, namely in a meta-search environment on the Web. It is important to mention that such knowledge can be reused in many other domains. However, the focus of this thesis will be to explain how the retrieval effectiveness of traditional meta-search techniques can be improved by reusing semantic knowledge represented by multiple merged and dynamically enriched ontologies. The basic idea is that merged domain-specific and other general-purpose ontologies can be used to tackle the problem of ambiguous words (in the user’s query, as well as in the returned search results) and hence improve the retrieval effectiveness of the meta-search engine. We present a detailed description of the proposed search model and discuss its main advantages compared to other classical keyword-based search models. In addition, we experimentally show that reusing ontological background knowledge has significantly improved the retrieval effectiveness of the meta-search engine that we developed to test our proposed search model.