-
Logo
Inspiration
Our group joined with a mission statement: how can we make mental health services accessible?
What it does
MindMap is a web application in which users write daily journal entries. It features a trained machine learning model that can perform sentiment analysis on the journal entries and will report the top two identified moods of the entry and apply a color to the user's calendar based off the emotion classification.
How we built it
There are two models that a developer can switch between should they choose; one is a linear SVC trained on a dataset consisting of utterances that have a mood classified to them and the second is a pretrained BERT transformers model for sentiment analysis. The current code is using the pretrained model because it is faster and more portable. The backend is built using Python Flask with an SQLite database to keep track of journal entries. There are routes from the backend to the HTML/CSS frontend.
Challenges we ran into
Trying to train a strong model took a large portion of time before we decided to use a pretrained model instead. The largest challenge we ran into was building the calendar, more specifically we were unable to get the mood scores provided by the model to match up with the dates of the post to color in the calendar.
Accomplishments that we're proud of
Getting the model to predict reliable scores, setting up a backend and frontend.
What we learned
We planned out too many features and experienced "feature creep". In the future, we should try to plan out a more feasible task given everyone's skillset.
What's next for MindMap
An original idea for MindMap was to restrict users to posting one journal entry a day, we were unable to implement this in time, so this would be the first next step for MindMap. Finishing features such as coloring the calendar with a gradient to signify the multiple moods classified throughout the journal entry. A secure login and account system.
Log in or sign up for Devpost to join the conversation.