Summarizing user action sequences with data analysis

• Main Author: Low, Bertrand Yilun
• Language: English
• Published: University of British Columbia 2011
• Online Access: http://hdl.handle.net/2429/31983

In the never-ending pursuit to enhance user interaction with computer systems, generating useful summaries can be highly beneficial. Summaries can provide the analyst with a map of past user behaviour that can aid in predicting future user actions. Tasks, such as forwarding an email or publishing and updating webpages, are composed of many individual user actions - as such, we view each task as a tree with the leaves of the tree representing the goal(s) of the task. We present a framework for modeling user actions as Navigation Trees and summarizing them into Summary Trees. The Summary Trees can be used to help streamline subsequent user action by acting as a guideline to semi-automate tasks. Using the concept of coverage and varying the number of attributes considered, we show how the quality of a Summary Tree can be adjusted. We also discuss five algorithms that approach summarization differently, compare their advantages and disadvantages, and provide an experimental study to empirically examine their individual characteristics. === Science, Faculty of === Computer Science, Department of === Graduate