Hyperstructure-Based Search Methods for the World Wide Web

This thesis presents several hyperstructure-based Web search methods and a prototype system that is designed to implement the methods. Given the context of hyperlink structural and semantic information that is representable with new Web standards, this thesis is an effort to answer the open question of how to efficiently make use of such information for searching the Web and filtering and retrieving relevant information. The hyperstructure-based approach taken in this thesis is an extension to the traditional structure-based search method, which mainly handles hierarchical structures (composed by non-linking mechanisms) in structured documents (e.g., XML). In addition to such hierarchical structures, this approach can also handle both hierarchical and non-hierarchical structures composed by linking mechanisms. Compared to other link-based approaches that largely take into account the quantity of links in their search methods, this approach also makes use of the semantic information in links and link-based structures. It is in line with the trend of Web development with regard to capturing rich structural and semantic information and thereby capitalizing on the potential of new search methods. The hyperstructure-based search methods presented in this thesis can be applied to improve the search quality on the Web as the Web evolves from a poorly structured to a more structured, semantic-rich network. More concretely, by making use of hypertext composites and contexts, the search results can be more specific with respect to users’ information needs, and additionally, the users’ efforts to interpret the search results can be reduced. Presenting structured search results based on hypertext composites as inter-linked nodes/pages rather than separate nodes/pages helps users understand the retrieved information better. By making use of semantic information in hyperstructures (e.g., types of links and nodes), better filters can be developed for selecting and ranking the Web pages retrieved by search systems. These pages can be either intermediate information for further processing or final search results presented to users. By making use of domain models, domain-specific structure-based search methods can be developed, which may generate better results than general search methods that do not understand the domain-specific information