Abstract

There are several issues at play within Enterprise Search, a mayor one is that enterprises are finding it difficult to retrieve their data using current Enterprise Search (ES) solutions.
This thesis work is a step towards improving this. By designing, implementing and evaluating a system for Enterprise Information Retrieval (IR), with an improved search result presentation technique, which assists users during search tasks. This was accomplished by exploring current research within the IR domain, then researching current ES systems and search result presentation techniques.
This research resulted in finding two forms of document summaries, namely a textual document summary technique: Top Ranking Sentences (TRS) and a visual document summary technique: Thumbnails. These document summaries have been designed to support a user during information seeking activity. A cognitive model for IR was also designed to get a better understanding of the information seeking process.
The system developed was named Basora Enterprise Search (BES) and was developed using an agile software development approach. It incorporates the TRS and Thumbnail into its search result presentation technique.
The BES prototype was put through both a performance and user evaluation. The first test indicated that BES performs more or less equal to the Commercial Enterprise Search Solution IBM Omnifind 8.5, while providing the user with two extra forms of document summaries.
The user evaluation focussed on evaluation of the effectiveness of the new search result presentation technique. The results of the user evaluation show that there are various search tasks where the addition of these summary elements has a positive effect on relevance assessment and query reformulation.
This research indicates that the BES system actually helps a user assess the relevance of a document, minimizing the amount of documents that need to be opened before the user finds the desired one. It also indicates that the additional visual and textual document summaries assist the user when reformulating a search query, decreasing the time it takes a user to complete a search task.