Charalampos Kokkalis, Giannis Tyrovolas, Ioannis Stamoulis
Our hack is an ambitious way of connecting moods with songs. As people who enjoy music we found it very common that algorithms entrench our pre-existing preferences when it comes to music. We wanted to create a new way to explore new music. That's why we created DJ Mood. We use natural language processing to identify the user's emotions and create a personalized playlist for each user.
Creating this hack was great fun and we learned a lot of new things in the process.
The high-level plan was the following:
0a. Analyze features of a large dataset of songs 0b. Create a website and a simple interface for the user
- Use input data to create a mood vector
- Associate our mood vector with each song by training appropriate weights
- Create a link with the user's new playlist
Analyse features of songs
We chose as our dataset Spotify's top 100 playlists. We did this because it includes a diverse set of musical genres and moods. With playlists from RapCaviar to Mellow Piano we believe a huge range was covered.
We used Python's Spotipy API to gather this data.