Inspiration
As a junior, I was released from having to live on campus this year, but that came with having to do extensive research of listings in the area. One source of reviews that I found quite helpful was Reddit, because it feels more transparent, and more MSU-oriented in terms of what the average student would need. I came up with this idea of an extension to make the lookup more convenient, as well as planning more useful features from those data.
What it does
It would detect if the user is looking at a listing on Apartments.com for example, extracting the name and getting data from the backend.
How we built it
The backend queries Reddit and Google's APIs to gather relevant threads, runs it through an optional analysis by OpenAI's API, then pre-renders it on the server and caches the result for fast retrieval. I made use of Redis as a cache and a job queue, and MongoDB for the persistent listing data.
The front end is a simple jQuery/Bootstrap interface.
Challenges we ran into
A major challenge was actually filtering out useful threads from Reddit. Reddit's API is not good with token matching, and Google is way better but has a rate limit. I am exploring various NLP approaches to improve the actual search queries as well as filtering out irrelevant results.
What we learned
While making this project, I applied new approaches to development that I had learned, like caching, scheduling, and rendering on the server. I achieved great results with them and will definitely incorporate those concepts into future projects.
Log in or sign up for Devpost to join the conversation.