Fantasy Football Sentiment Analyzer

I love fantasy football, but I am unfortunately not very good at it. I have played 2 seasons, and have yet to make the playoffs. I felt that surely I could take advantage of my Computer Science degree to turn my team around this season. As my record shows, I am by no means an incredible football analyst, but I have seen videos from people who do this for jobs predicting which players will Boom and will players will Bust every season. And a lot of these analysts are involved in podcasts. As a type of long-form content, podcasts offer a deeper form of insight than I usually get from the shorter videos I see. But who has the time to listen to all these podcasts? And so I decided to build a tool that could do it for me.

I built the website itself using Next.js and Tailwind CSS, and set-up a simple Flask API so that the frontend could communicate with the backend. I wanted to mimic how MonkeyType.com allows the user to seemingly type into the website rather than using a clear form of input, and I am happy with how my rendition came out.

And in the backend, I built an end-to-end NLP pipeline that can take in an audio fie of whatever podcast I want to learn from, and output every NFL player mentioned in the podcast, along with some sentiment scores indicating what the experts really thought of the player. However, transcription has proven to be a very resource intensive process, and while I could optimize transcription, the accuracy trade-offs often result in significantly worse sentiment insights. So I decided to start the pipeline assuming the user provides a generated transcript.