Summary
People tend to listen to songs in particular groupings that aren’t always obvious from the artist or genre labels. The authors studied a huge amount of listening data to see if they could find hidden patterns about how songs are experienced together. They used a machine learning model to create a map of songs based on co-listening sessions, then developed a way to remove the influence of artist identity to find deeper connections. Their method revealed groups of songs from different artists that share similar styles or moods, showing there’s more to music similarity than just artist or genre. This helps reveal how people actually experience music in a more personal, subtle way.
What this means in practice
- •For music streaming services: Design song recommendation algorithms that capture deeper listener experience beyond artist and genre labels using session co-listening data.$Commercial implications: This paper enables improved music recommendation products by recovering experiential song similarities unseen from existing metadata, enhancing user engagement.
- •For playlist curators: Identify song clusters that reflect listener experiences across artists and genres to create more coherent and varied playlist themes.
Abstract
This report presents results from Project Qualia, an ongoing effort to determine whether experiential similarity between songs, a structure not captured by genre or metadata taxonomies, can be recovered from real listening behavior. We constructed a large-scale dataset of listening sessions, comprising 1.29 billion scrobbles collected from 9,396 users via the Last.fm API and reduced through a preprocessing pipeline to 531.6 million training scrobbles across 28.6 million sessions. On this corpus, we trained a skip-gram Word2Vec model (Song2Vec), treating each session as a sentence and each track as a token. As anticipated, the resulting embedding space was dominated by artist identity, a consequence of single-artist runs within sessions. To test for a subtler, artist-independent signal, we developed an artist-residual procedure: subtracting each artist's centroid from its tracks' embeddings and evaluating whether the remainder retained structure. Mean cross-artist cosine similarity fell from 0.2487 in raw embedding space to 0.0005 in residual space, yet 4,577 cross-artist track pairs retained cosine similarity $\ge 0.70$ in residual space, forming coherent genre- and era-based clusters, including trip-hop, 1990s grunge, 2020 mainstream pop, and cross-composer classical piano pairs at cosine similarity up to 0.95. These results confirm that the training data contains experiential structure independent of artist identity, establishing an empirical basis for an architecture designed to learn this experiential layer directly.