Identifying search patterns in an image‐based digital library
Hyejung Han, Dietmar Wolfram · Proceedings of the American Society for Information Science and Technology · 2014
ABSTRACT Three months of clickstream transaction log data for an image‐based digital library were analyzed for patterns in user session behavior. After extensive log cleaning of the data files, k‐means cluster analysis was applied using selected session‐related statistics. The outcomes reveal largely uniform session behaviors (97.37%) consisting of a limited number of actions, with a higher number of queries than other session types, but little browsing of individual results. Given the specialized nature of the digital collections within the studied system, most searchers appear to engage in purposeful, directed searching. The remaining session clusters represent higher degrees of browsing over a longer time with fewer queries per session. The session behaviors for this environment are markedly different than those revealed by transaction log studies of other information retrieval environments.