Association mining of search tags in PubMed search sessions

Background: Previous studies have shown that use of search tags in PubMed can significantly improve the performance of information retrieval. The objective of this study was to discover associations among search tags in typical PubMed search sessions. Methods: We performed session segmentation on a full-day PubMed query log, identified the search tags within those sessions, and applied association mining to identify strong associations of search tags. Results: A total of eight maximal frequent-itemsets (i.e. search tags) and 34 strong association rules from these itemsets were discovered. We also estimated that the query refinement occurs frequently (i.e. one query per minute on average) for any session length. Conclusions: The association rules consisting of PubMed search tags can be used to develop an interactive and intelligent PubMed search interface so that the users can build the search query using proper search tags and reduce the frequency of query refinement.