Mental Health on Tufts Sidechat

Inspiration

I have used Sidechat, primarily when I was a Freshmen. People speak freely on it, and people speak a lot on it. I realized that much of student life and student thought is actually is specifically on the platform! Inspired by my friend Matt Sheng, he thought that

What it does

It is a pipeline to ingest data from Sidechat and create powerful accurate embeddings use via the Tufts HPC. Then we can do downstream analysis on those embeddings, as well as using vote count and time data to both track general emotional trends as well as isolate "crisis posts."

How we built it

I built the scraper a few months ago myself. I used the Tufts HPC to initialize and create the embeddings and do emotional classification. Then I used python and standard libraries on my own computer to do the further analysis.

Challenges we ran into

HPC was having issues, kept on allocating a node that would not build my container! Just had to switch nodes, but took a lot of time!

Accomplishments that we're proud of

I am happy that we created embeddings for 212,000 students for free and so fast! The HPC was finicky. But honestly, I am truly the most proud of our analysis and how insightful and create the dataset I created was.

What we learned

LOOK AT MY PRESENTATION, I RUNNING OUT OF TIME. BUT WE FOUND WHEN PEOPLE EXACTLY ARE HACING CRISISES AND HOW EMOTION CHANGES OVER WEEKS AND YEARS AND OTHER COOL SHIT.