Integration of search theories and evidential analysis to Web-wide Discovery of information for decision support

The main contribution of this research is that it addresses the issues associated with traditional information gathering and presents a novel semantic approach method to Web-based discovery of previously unknown intelligence for effective decision making. Itprovides a comprehensive theoretical background to the proposed solution together with a demonstration of the effectiveness of the method from results of the experiments, showing how the quality of collected information can be significantly enhanced by previously unknown information derived from the available known facts. The quality of decisions made in business and government relates directly to the quality of the information used to formulate the decision. This information may be retrieved from an organisation’s knowledge base (Intranet) or from the World Wide Web. The purpose of this thesis is to investigate the specifics of information gathering from these sources. It has studied a number of search techniques that rely on statistical and semantic analysis of unstructured information, and identified benefits and limitations of these techniques. It was concluded that enterprise search technologies can efficiently manipulate Intranet held information, but require complex processing of large amount of textual information, which is not feasible and scalable when applied to the Web. Based upon the search methods investigations, this thesis introduces a new semantic Web-based search method that automates the correlation of topic-related content for discovery of hitherto unknown information from disparate and widely diverse Web-sources. This method is in contrast to traditional search methods that are constrained to specific or narrowly defined topics. It addresses the three key aspects of the information: semantic closeness to search topic, information completeness, and quality. The method is based on algorithms from Natural Language Processing combined with techniques adapted from grounded theory and Dempster-Shafer theory to significantly enhance the discovery of topic related Web-sourced intelligence. This thesis also describes the development of the new search solution by showing the integration of the mathematical methods used as well as the development of the working model. Real-world experiments demonstrate the effectiveness of the model with supporting performance analysis, showing that the quality of the extracted content is significantly enhanced comparing to the traditional Web-search approaches