Beyond actions: Exploring the discovery of tactics from user logs

摘要: A novel semi-automatic log analysis method to identify user search tactics.Unites empirical and conceptual information seeking behaviour models.Makes across systems of different nature design comparable. Search has become a common practice gain insights into behaviour: it helps an understanding needs preferences, as well insight how system supports such needs. Currently, is typically focused on low-level actions, i.e. logged events issued queries clicked results, often only selection are analysed. However, types may differ widely from interface interface, making comparison between difficult. Further, the interpretation meaning subsequent lead conclusions out context- e.g. statistics observed query reformulations be influenced by existence relevance feedback component. Alternatively, in lab studies activities can analysed at higher level, tactics strategies, abstracted away detailed implementation. Unfortunately, until now required manual codings that map higher-level interpretations have prevented large-scale use this type analysis. In paper, we propose new for analysing logs (semi-)automatically identifying events, allowing comparable systems. addition, resulting tactical level reduce potential issues surrounding need actions We validate efficiency effectiveness proposed tactic identification using two reference natures: product video system. With identified tactics, perform series analyses terms entropy rate sequences, demonstrating allows comparisons behaviours design. This provides not achievable with traditional Display Omitted