deScribe

Inspiration

Mental wellness and awareness is a key part of staying healthy. It's important to reflect and check in with yourself, identifying the things that make you sad, happy, angry, and anything in between. However, in an age where technology dominates our lives, we don't always make time to reflect.

What it does

deScribe to allow people to express their emotions and reflect on their day by talking with Kiwi, a Google Assistant chat bot. Kiwi will ask you questions about your day, and through IBM Watson's sentiment analysis, we'll provide you with a report of your feelings, and why you may be feeling a certain way.

We record all of your sessions in an online web app, and provide you insight on your conversation – consider it journalling, without the pen and paper.

Over time, as you continue to talk with Kiwi, you can see your progress on a dashboard. Responses are analyzed for keywords, used to determine what made you sad, glad, or downright mad.

How we built it

We built Actions for Google Home that consisted of a trained model to understand the emotions the user is feeling based on what the user says. Once the emotions were parsed, the model would choose an appropriate response for the user, similar to a real conversation. Using Google Cloud Functions, we created webhooks to listen to user response and load data into Firebase. We then run sentiment analysis on the user’s journal entry to determine key words, overall sentiment, and emotions. Our companion web application extracts the results of the sentiment analysis and displays the information on a dashboard.

Challenges we ran into

Training our model for Google Home to facilitate a natural and supportive conversation with the user
Connecting the individual parts of our application

Accomplishments that we're proud of

Setting up a webhook with Google Cloud Functions that sends user response data to the server

What we learned

Planning is crucial to the success of a project, especially with a large one like deScribe where there are many components that need to be tied together