User evaluation of textual results clustering for web search

Purpose: Clustering web search results into dynamic clusters and hierarchies provides a promising way to alleviate the overabundance of information typically found in ranked list search engines. This study seeks to investigate the usefulness of clustering textual results in web search by analysing the search performance and users' satisfaction levels with and without the aid of clusters and hierarchies. Design/methodology/approach: This study utilises two evaluation metrics. One is a usability test of clustering interfaces measured by users' search performances; the other is a comprehension test measured by users' satisfaction levels. Various methods were used to support the two tests, including experiments, observations, questionnaires, interviews, and search log analysis. Findings: The results showed that there was no significant difference between the ranked list and clustering interfaces, although participants searched slightly faster, retrieved a larger number of relevant pages, and were more satisfied when using the ranked list interface without clustering. Even so, the clustering interface offers opportunities for diversified searching. Moreover, the repetitive ratio of relevant results found by each participant was low. Other advantages of the clustering interface are that it highlights important concepts and offers richer contexts for exploring, learning and discovering related concepts; however, it may induce a certain amount of anxiety about missing or losing important information. Originality/value: The evaluation of a clustering interface is rather difficult, particularly in the context of the web search environment, which is used by a large heterogeneous user population for a wide variety of tasks. The study employed multiple data collection methods and in particular designed a combination of usability and comprehension tests to offer preliminary results on users' evaluation of real-world clustering search interfaces. The results may extend the understanding of search characteristics with a cluster-based web search engine, and could be used as a vehicle for further discussion of user evaluation research into this area.