Methodology for non-systematic information seeking on the web

In an ideal world, once a user issues a query, the required information is found in one of the first few documents returned. However, search engines often fail to return documents with required information. This dissertation focuses on what happens next, after a search engine user decides that required information is not in the returned document set. With the existing search engines, the only action users can perform is to revise the original query and try again. If the original query and the revised query are topically related, their corresponding result sets will also be related and even contain a large degree of overlap. By repeating this process, a user explores the neighborhood of documents related to a topic of interest.; This dissertation proposes more direct tools for users to explore the document space based on tracing the hyperlinks in web documents. An interactive search model comprising search cycles of querying, navigation, scanning, and comprehension is proposed. The aim is to facilitate the understanding of information seeking within the Internet environment by integrating all aspects of information seeking. A number of characteristics of information seeking on the web can be identified relating to the research on browsing strategies. Browsing strategies are primarily concerned with the organization of search tactics and moves in a way to complete a search effectively.; This project reports on the software design issues of the proposed browsing strategies, focusing on navigation tools that make up a user interface responsive to the topic narrowing procedure. "Topic Narrowing" is a mental procedure that is often used by people when searching for information. Browsing capabilities provided in the proposed model include backward surfing, neighborhood visualization, multi-level comprehension with spatial document visualization, and zooming. The effectiveness of the search model is analyzed independently for each browsing strategy. From a cognitive viewpoint, local and global coherence, cognitive overload, and the use of representative images are among the factors analyzed.