This week was focused on the intent recognition of our software. More specifically, we scheduled a meeting with Dr. Juan Gilbert, an expert in intent recognition and the department chair at UF in computer science. This meeting was exceedingly beneficial for the team as we have been struggling to make the intent recognition module optimal for our presentation at the end of the semester.

During this meeting, we were told that our original intent recognition, a time-based approach, is a good start – but can be better! This surprised us as we thought the time-based approach will be completely thrown out the window and replaced with a machine learning element. While this is what we thought Dr. Gilbert would suggest, we honestly didn’t know how to implement the machine learning aspect since our company wouldn’t be able to provide any data for the machine to learn from. Therefore, hearing that the time-based approach would work was amazing! Some improvements as suggested by Dr. Gilbert, are as follows: 1. Have the time-based threshold adapt with the user 2. The time-based threshold should also change depending on the element that the user is currently on 3. Identify if the user went back to an element, and if so, that will also be an indicator of intent 4. Identify if the user selected/repeated a certain section, and if so, this is also an indicator of intent.

Overall, the team is happy to have an intent recognition module be fixed with an easier solution than previously realized! This upcoming week we are hoping to start implementing these suggestions and within the next couple of weeks we hope to then move on to beta testing to get feedback on the intent recognition side of our software!