Inspiration
We initially had in mind a project that could benefit people during the everlasting lockdown in Quebec. We pinpointed certain issues people might be facing during these times, and concluded that loneliness is the most impactful. Knowing that therapists are sometimes expensive and overbooked, we believe that having a buddy to chat with when you want, as soon as you need, would be most beneficial for people.
What it does
Lockdown buddy instantly responds to your messages and always has something interesting to say. It can answer basic math problems, and can always change the subject to something random if it doesn't understand your messages. This way, no matter how you're feeling, you're guaranteed to get lost in the conversation with this spontaneous robot.
How we built it
The team started by setting up the WebSocket server for instant real-time messaging. Next, we built out the frontend side of things with EJS and HTML. After that, we implemented a natural language processing machine learning model to help our chatbot provide instant and accurate responses to all the user's questions. Finally, we connected the backend with durable storage in the form of cockroachDB to replicate all our user's data across nodes. This means any user can hop right back into a conversation with Lockdown Buddy.
Challenges we ran into
The team ran into some issues with the frontend side of things, since we weren't fully familiar with HTML and CSS. The centering and alignment of text messages between the user and Lockdown buddy was rather complicated initially, but the team was able to quickly find new techniques to improve the UI.
Accomplishments that we're proud of
Several members were completely unfamiliar with proper frontend design, and the team is proud of the final result that truly resembles a chatting app such as iMessage. Although it may be simple for others, this was a steep learning curve for us and we are happy with the outcome.
What we learned
The team developed a new understanding for cockroachDB, basic machine learning, frontend development techniques, natural language processing and data processing.
What's next for Lockdown Buddy
Definitely improving it's speech training, so it can better understand how certain people are currently feeling, and how to best respond. It is currently limited in its knowledge, and may not know exactly what to say at all times. More AI training is crucial and would make Lockdown buddy a lot more fun to chat with.
Log in or sign up for Devpost to join the conversation.