list from [[]] and !outgoing([[]])🧾 Project Description
A project aiming to recommend relevant songs to an individual, either through connecting Spotify or finding similar songs
📂 Project Logs
08-30-2024 16:13 - I’m currently looking into Vectorization of song attributes, but I’ve also heard that maybe a KNN algorithm could be helpful.
09-02-2024 16:27 - Creating these pages:
09-17-2024 18:13 - Back from trip, reading up on all the leads I had. Right now a KNN approach looks to be the clear path forwards. I wanted to also do album recommendation, but that sounds really hard. I’m not sure what approach to take with that, I think I’ll tackle it after getting the song recommender off the ground. My lead with that is giving each album an average rating in each category by combining the song attributes, then also recording variance in each category, and maybe creating derived features.
03-22/2024 9:43 - Done with Winter quarter, that was tiring. Most interesting lead now is cosine.club, an interesting MRS that actually uses audio qualities of songs to recommend similar songs. This is better than a hack I was planning of just looking up songs in some music genre finder.
- Do a Project-Analysis on cosine.club
🗣️ Blurt
I’m very excited about the possibility of doing something interesting. It was a bit of a wake-up moment to see how boring the site was to JD, I’m not making another boring app. Everything must do something useful, or something genuinely cool.
🏢 Existing
🔨 Todo
Biggest to-do is research on common techniques. First lead is the web app a guy did when connected to Spotify app. I have a suspicion that he just used the Spotify API recommendations, but maybe there’s a cooler idea under the hood.