/Peritia

Chatbot created for the Convergence.tech’s Digital Inclusion Challenge 2020.

Primary LanguageASP.NET

Peritia

Chatbot made for the Convergence.tech’s Digital Inclusion Challenge 2020. The Challenge was one week long.

The project was to create a chatbot that would assist students in developing productive habits. The chatbot would give helpful strategies to the user to assist them in developing good habits.

The original plan was for the chatbot to be coded in Python and to incorporate sentiment analysis which would be used for mental health purposes. A supervised learning model was utilized to generate responses. However, this method had many issues. The accuracy of the sentiment analysis was not very high, and at the time I had very little experience with finding strong datasets and creating models from the data. The chatbot also needed to be ran on another computer but since Python was being used, a computer would need to have Python on their computer to run the program. There were solutions to these issues, but the timespan was short and thus Python was abandoned and a new method was used instead. The code from the Python implementation is still in this repository.

The project was then coded in C# using Unity. I had more experience with this platform and so I was quickly able to make up for the lost time spent on the Python implementation. However, sentiment analysis was abandoned and a rule-based chatbot was used instead of using a supervised learning model, due to the time constraint. The GUI was quickly built using Unity's drag and drop feature. The application was tested on an Android device as well as on a computer. With the new option of building for Android devices came a new feature for the chatbot, creating notifications. The chatbot would create notifications at certain times determined by the user in-app to help with their productivity. However, when tested on the Android device, the notification feature would not work. This was most likely due to the test device being an older device and perhaps the notification feature was not compatible with the device. The app was only tested on one device, thus this issue was not further pressed. Some features of the application were not finished in tiem for the deadline, but the entire application did not need to be shown, so this was not an issue. Currently, the application is still not finished. I hope to finish it soon, whenever I next have time.

The overall project made it to the Top 30 Submissions for the Challenge.