Effectiveness of real-time query expansion in a library catalog

Abstract

Query expansion (QE) is a potentially effective technique to help searchers formulate or improve their search. In January 2010, the University of North Carolina at Chapel Hill (UNC) Library made an effort to improve user interactions with authority data in its online catalog. It introduced a realtime query expansion feature for the catalog interface that would provide real-time suggestions based on what the user had typed in the search box. QE is not a new technique, and has been widely used in web search interfaces. It has rarely been adopted by library catalogs though. From the libraries perspective, populating the query expansion feature with authority data provides a balance between traditional library use of authority records and the expectations users have from their use of Web search engines. By analyzing the search logs from the library server, we investigate the impact of this feature on users' search behavior. Two sampling periods of logs were chosen: April 2009 (before the feature's launch) and April 2010 (after the feature's launch). Totally, we have analyzed approximately 200M log data with approximately 800,000 useful log records. Results showed that users take the suggestions 20.26% of all searches. There was a significant increase in the rate of subject searches. The library catalog interface supporting dynamic query expansion also leads to longer queries, less query iterations per search, and fewer result items viewed per search. In addition, search failure rate has seen a great drop. However, results also imply some issues may exist with the use of query expansion, necessitating proper handling of the search expansion interface. Our findings have implications for how query expansions should be offered to library catalogs to increase the use of library's collection.