Exploro search agent

Abstract

The World Wide Web is a vast maze of information. Navigating through this huge search space successfully is not a trivial task regardless of the user's search skill. In this dissertation we describe Exploro, a browser-side software agent that aids the user's textual and pictorial information retrieval needs. Exploro aids the user in three ways; automatic query generation, image search and automated bookmarking. Automatic query generation caters for three common search needs. First, search of a particular term of interest in the context of the page. Second, search of similar pages to the one currently viewed. Third, automated queries for link recommendation of topics present in the current web page. Image search aims at gathering images similar to the query through a hybrid two step approach; using the surrounding text a query is generated automatically and dispatched to an image search engine, the resulting thumbnails are then compared for similarity with the original query image and ranked. Automated bookmarking identifies the most appropriate bookmark folder the web page to be bookmarked should be placed. Exploro also learns a user profile unobtrusively by analysing the content of the web pages viewed, building a "hierarchical continuum of general to specific user interests" [25] and is used in query generation and automated bookmarking. This user profile is then evaluated at its ability to categorise documents into the correct bookmark folder and results indicate a high success rate when the bookmark folders contain related web pages.