This project is designed to get your feet wet with some real-world data. Here are a couple of suggested project options:
Research-style project: train and evaluate some models on an NLP task of your choice. This will involve finding some data, preprocessing it, training models, and doing some analysis. Feel free to use pretrained models.
Hackathon-style project: build a pipelined speech-to-speech translation system. This was the project from last year, and you can find more info here.
Something else! Do something that interests you.
You should do the project in pairs. You can find a partner on Piazza. We'll setup a time next week to talk briefly with each group on what they want to do.