Russian search engine giant Yandex Released on May 30th, the world'sThe largest music recommendation system open sourceDataset YambdaContains 4.79 billion pieces of anonymized user interaction data designed to help developers create smart music services.
Yandex collected in ten monthsNearly 28 million monthly Yandex Music subscribersThe data, specifically the user4.79 billion interactions with 9.39 million songs, the dataset includes key feedback from listeners on the goodness or badness of the song, and all interactions areWith timestampto improve accuracy.
Yambda at Hugging Face Three sizes of data sets are availableFor free download: Yambda-5B (behavior from 1 million users), Yambda-500M (100,000 users) and Yambda-50M (10,000 users). Among the sizeslargest 5B dataset requires a minimum of 85 GB of storage space.
The dataset contains information about the preferences of music listeners, theStored in Apache Parquet formatThe company's music recommendations can be used for research purposes or to develop AI music recommendation features similar to those offered by streaming services such as Spotify.
Streaming services such as Spotify and Tidal don't typically release the code or models for their music recommendation algorithms because the ability to recommend a listener's favorite song is seen as a trade secret to their success, 1AI understands.