As UofM students, we constantly found ourselves lost in a maze of offices, websites, and departments not knowing who to contact for advising appointments, or campus resources. We'd get bounced from one office to another, wasting time and energy. We built BisonBot to be the guide we wished we had on day one.

BisonBot is a University of Manitoba AI Advisor who accurately answers questions such as “What is the deadline to drop a winter course?” with the correct information from U of M’s academic calendar or student guides.

We used a framework called Angular on the front-end and for the back end we trained rasa bot and used its api to send message and get responses.

Gathering and structuring accurate UofM-specific data was the biggest challenge because a lot of information is scattered across different websites. We also had challenges with hosting the working api locally which consumed time.

Even with a lot of bugs and errors, we still had the website up and running. We learnt a lot of things about how chat bot works and how we can train them. Another challenge was to use rasa because it only runs on python version 10.

We learnt how to integrate an AI chatbot and how to train it on our own system. This chatbot can totally be ran offline if needed it can even be enhanced with modern api like open ai, claude etc. We also learnt how to connect the chatbot to the front-end using apis.

We can further train the bot to provide more accurate information and also expand it to be an advisor for other faculties and not just limited to science. We are also planning to add user login and persist data so that users can visit back and see the responses.

Built With

Share this project:

Updates