Check out our slides here: shorturl.at/fCLR4
In a digital age increasingly dictated by streaming services’ editorial teams and algorithms, we want to bring the emotion back to the music discovery experience. Your playlists have become an echo chamber of your own past preferences and the short, chorus-led song mold artists now follow to maximize playlist performance, giving music streaming platforms much of the control over who succeeds in this industry and what types of music users can easily access. We want to democratize the process of finding unique, emotion-driven sound and bring your personal emotions back to the forefront of your listening experience.
What It Does
Thus Maestro was born, a chat bot that integrates with virtually any messaging service and enables a user to send an emoji corresponding to how they’re feeling and receive a personalized new song recommendation through YouTube.
How We Built It
To achieve this maximized integrability, we used Gupshup to build the Messenger chatbot and developed the back-end on a Python Flask-JSON localhost server exposed on Ngrok. The AI sentiment analysis is performed a public, large Kaggle dataset scraping tweets with emojis. Given the sentiment an emoji conveys, we then match it to a corresponding YouTube playlist, and scrape the playlist for a suitable song track.
Challenges We Ran Into
The FB Messenger API was hard to access - we wasted some time initially trying to gain access to build on top of it & to integrate it with AWS. That’s when we switched over to GupShop.io. However, it was extremely clunky. It slowed us down massively and required a lot of duplicative work.
We also struggled to combine our front end and back end in one continuous integration pipeline. After receiving a tip from a mentor on what resources we could use to achieve this, it took massive efforts to integrate it all together.
Accomplishments We're Proud Of
A working Messenger bot that can be used by anyone, at anytime! These are emoji-song recommendations that are actually GOOD :)
What We Learned
Through our hack, we learned to perform sentiment analysis on not just text but emojis as well. We also learned how to integrate back-end and front-end -- especially with hosting a server and exposing it on a URL, as well as setting up a Messenger chatbot most smoothly.
What's Next For Maestro
We're very excited to develop Maestro further, including building a feedback loop so that users can tell us if they like their song, adding personalization so that music recommendations are tailored to the user, and catering to more streaming platforms so users can listen on Spotify/Apple Music/Youtube/etc. We'd also like to port to other platforms so that we achieve integration with Whatsapp, SMS, Instagram DMs, and others, as well as build on human curators to better select music.