| Abstract |
Although most information needs have exploratory nature, current web search engines do not support these needs sufficiently. The objective of this thesis is the development of effective and efficient services for fulfilling such needs. Specifically, this thesis proposes the extension of the "traditional" interaction scheme (query-and-ranked answer) with exploration services based on dynamic faceted taxonomies. The advantages of this interaction scheme is that it bridges the gap between querying and browsing, it provides an overview of the information space, and allows the user to reduce the size of the information space in a simple, flexible and gradual manner.
To support this model over a Web Search Engine (WSE), we studied various architectures and implementation approaches and subsequently we designed and implemented a general purpose API and a Web-based GUI interface. Furthermore, we examined several methods for exploiting the metadata that are available at the index of a WSE plus those that can be dynamically generated (e.g. from applying real-time clustering results techniques or other data mining techniques). The above techniques have been applied and evaluated over the Mitos WSE, making it the first WSE that supports this model. The performance of the proposed techniques was evaluated experimentally, while various variants of the model (regarding the choice of facets) were evaluated by users comparatively. The evaluation showed that users can use this model of interaction without any kind of training, and can use it effectively.
Finally, the thesis proposes and analyzes a method for scaling up these services that relies on CTCA (Compound Term Composition Algebra) and on special indices and algorithms. The experimental evaluation showed that the proposed approach is more efficient than the existing approaches.
|
Efficient faceted exploration services for big volumes of Information
