Transition Discovery of Sequential Behaviors in Email Application Usage Using Hidden Markov Models

Requirements monitors provide high-level feedback on software usage in real-time. Herein, we show how low-level monitoring can identify behavioral transitions that can be interpreted as learning transitions by a post-clinical team. One monitoring technique is to apply stochastic modeling to the software´s event stream. Herein, we show how dynamically generated hidden Markov models (HMMs) characterize sequence patterns in a software´s user-interface event-stream. We show how this is used to dynamically model a user´s usage of an emailing application. By differencing the resulting sequence of generated HMMs, the technique can identify transitions in software usage. This is important for identifying usage transitions, which occur with user learning. Herein, we show how the approach applies to monitoring an email application that has been simplified for users having cognitive impairments. The identified transitions provide the post-clinical team feedback on a user´s emailing progress. The team then uses the feedback to make adjustments to the emailing environment to further aid learning.