Web User Interaction Mining from Touch-Enabled Mobile Devices

Web services that thrive on mining user interaction data such as search engines can currently track clicks and mouse cursor activity on their Web pages. Cursor interaction mining has been shown to assist in user modeling and search result relevance, and is becoming another source of rich information that data scientists and search engineers can tap into. Due to the growing popularity of touch-enabled mobile devices, search systems may turn to tracking touch interactions in place of cursor interactions. However, unlike cursor interactions, touch interactions are difficult to record reliably and their coordinates have not been shown to relate to regions of user interest. A better approach may be to track the viewport coordinates instead, which the user must manipulate to view the content on a mobile device. These recorded viewport coordinates can potentially reveal what regions of the page interest users and to what degree. Using this information, search system can then improve the design of their pages or use this information in click models or learning to rank systems. In this position paper, we discuss some of the challenges faced in mining interaction data for new modes of interaction, and future research directions in this field.