SignalSense

SignalSense

Inspiration

A couple of years ago, someone close to us struggled with severe mental health challenges and ultimately tried taking his life. It was hard to understand how clear the signs were until the situation escalated. That experience revealed a crucial gap: early warning signs are often too subtle and missing in everyday conversations. We decided to build Signal Sense as a real solution for everyday wellbeing, allowing prevention before situations become critical.

What it does

Signal Sense analyzes screenshots of conversations (e.g, messages) and evaluates the emotional and psychological signals in them.

Users upload a screenshot of a conversation. Our system extracts and processes the text. We use NLP from sci-kit learn to convert text into numerical features, allowing a logistic regression model to detect patterns in language associated with different mental health states. The system outputs probability scores across multiple mental health indicators with one main classification based on the highest probability score. Based on classification, gives you advice to deal with your family/friend.

The goal is to provide early awareness of potential distress so that users can make more informed decisions about checking in or offering support.

How we built it

We built a system that takes a screenshot, extracts the text using OCR, converts that text into numerical features, and feeds it into a trained machine learning model to predict the likelihood of different mental health categories and overall risk level. We used lots of sci-kit-learn documentation and youtube videos for the NLP and logistic regression.

Challenges we ran into

Lack of communication sometimes and a lot of errors in connecting HTML to flask. Also on using our PIL image to text tools since it was rather new to us and we were documentation scraping.

Accomplishments that we're proud of

The finished product! Also the fact we were able to get image to text working. Its a really new thing to us.

What we learned

The importance of task delegation so that everyone has something to do, the importance of adding documentation within our code so others can understand.