Patterns and Sequences of Multiple Query Reformulations in Web Searching: A Preliminary Study

While some studies have investigated query reformulation in traditional online systems, there has been little understanding of how users reformulate their queries multiple times within search sessions on the Web. This paper reports on patterns and sequences of query reformulation based on query logs from a Web search engine. The data set contained only search sessions in which multiple query modifications were made. The analysis of data resulted in three facets of reformulation: content, format, and resource. Each facet was further categorized by 10 sub-facets. The results show that while most query reformulation involves content changes, about 15% of reformulation is related to format modifications. Six patterns of query reformulation emerged as a result of sequence analysis: specified reformulation, parallel reformulation, generalized reformulation, dynamic reformulation, format reformulation, and alternative reformulation. Each pattern is discussed with definitions and examples. The results indicate that both planned and situated aspects affect query reformulation in Web searching. The implications for new Web search engine tools and features are also discussed.