claire

Login page featuring simple UI
AI chatbot
Dashboard of analytics for family members and caregivers to monitor

Inspiration

Loneliness affects countless people and over time, it can have significant consequences on a person's mental health. One quarter of Canada's 65+ population live completely alone, which has been scientifically connected to very serious health risks. With the growing population of seniors, this problem only seems to be growing worse, and so we wanted to find a way to help both elderly citizens take care of themselves and their loved ones to take care of them.

What it does

Claire is an AI chatbot with a UX designed specifically for the less tech-savvy elderly population. It helps seniors to journal and self-reflect, both proven to have mental health benefits, through a simulated social experience. At the same time, it allows caregivers to stay up-to-date on the emotional wellbeing of the elderly. This is all done with natural language processing, used to identify the emotions associated with each conversation session.

How we built it

We used a React front-end served by a node.js back-end. Messages were sent to Google Cloud's natural language processing API, where we could identify emotions for recording and entities for enhancing the simulated conversation experience. Information on user activity and profiles are maintained in a Firebase database.

Challenges we ran into

We wanted to use speech-to-text so as to reach an even broader seniors' market, but we ran into technical difficulties with streaming audio from the browser in a consistent way. As a result, we chose simply to have a text-based conversation.

Accomplishments that we're proud of

Designing a convincing AI chatbot was the biggest challenge. We found that the bot would often miss contextual cues, and interpret responses incorrectly. Over the course of the project, we had to tweak how our bot responded and prompted conversation so that these lapses were minimized. Also, as developers, it was very difficult to design to the needs of a less-tech-saavy target audience. We had to make sure our application was intuitive enough for all users.

What we learned

We learned how to work with natural language processing to follow a conversation and respond appropriately to human input. As well, we got to further practise our technical skills by applying React, node.js, and Firebase to build a full-stack application.

What's next for claire

We want to implement an accurate speech-to-text and text-to-speech functionality. We think this is the natural next step to making our product more widely accessible.