Music Recommendation Chatbot Game

During summer 2020, I worked as a Software Engineer Intern at Nautilus Software Technologies, where I led a team of four in building a chatbot game that delivers Spotify song recommendations to users in the Nautilus Facebook Messenger chatbot platform. We first built an external Python-Flask application to crowdsource labels for audio data scraped with the Spotipy library. Then, we built a Javascript application in the chatbot platform to gather users’ music preferences and integrated it with an external Express.js application, which pipes preferences to a Python script running a modified k-nearest neighbors machine learning algorithm. During the first four rounds, the Python script returns two new songs for the user to choose between, and during the last round, it returns a recommended mini-playlist based on the taste vector learned for that user and the database of labeled songs.

• The audio-tagging application can be found here.

• The GitHub repositories for the audio tagging and playlist generation components of this project can be found here.