INTERNET BROWSING AND SEARCHING - USER EVALUATIONS OF CATEGORY MAP AND CONCEPT SPACE TECHNIQUES

The Internet provides an exceptional testbed for developing algorithms that can improve browsing and searching large information spaces, Bro wsing and searching tasks are susceptible to problems of information o verload and vocabulary differences. Much of the current research is ai med at the development and refinement of algorithms to improve browsin g and searching by addressing these problems. Our research was focused an discovering whether two of the algorithms our research group has d eveloped, a Kohonen algorithm category map for browsing, and an automa tically generated concept space algorithm for searching, can help impr ove browsing and/or searching the Internet. Our results indicate that a Kohonen self-organizing map (SOM)-based algorithm can successfully c ategorize a large and eclectic Internet Information space (the Enterta inment subcategory of Yahoo!) into manageable sub-spaces that users ca n successfully navigate to locate a homepage of interest Po them. The SOM algorithm worked best with browsing tasks that were very broad, an d in which subjects skipped around between categories. Subjects especi ally liked the visual and graphical aspects of the map. Subjects who t ried to do a directed search, and those that wanted to use the more fa miliar mental models (alphabetic or hierarchical organization) far bro wsing, found that the map did not work well. The results from the conc ept space experiment: were especially encouraging. There were no signi ficant differences among the precision measures far the set of documen ts identified by subject-suggested teams, thesaurus-suggested terms, a nd the combination of subject- and thesaurus-suggested terms. The reca ll measures indicated that the combination of subject-and thesaurus-su ggested terms exhibited significantly better recall than subject-sugge sted terms alone, Furthermore, analysis of the homepages indicated tha t there was limited overlap between the homepages retrieved by the sub ject-suggested and thesaurus-suggested terms. Since the retrieved home pages for the most part were different, this suggests that a user can enhance a keyword-based search by using an automatically generated con cept space. Subjects especially liked the level of control that they c ould exert over the search, and the fact that the terms suggested by t he thesaurus were ''real'' (i.e., originating in the homepages) and th erefore guaranteed to have retrieval success.