Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Text2Playlist: Generating Personalized Playlists from Text on Deezer

Mathieu Delcluze, Antoine Khoury, Clémence Vast, Valerio Arnaudo, Léa Briand, Walid Bendada, Thomas Bouabça

TL;DR

This work tackles the gap between narrow, item-specific search and broad, exploratory text requests on Deezer by introducing Text2Playlist, a stand-alone text-to-playlist generator. The approach combines LLM-based tag extraction, tag-driven track retrieval, collaborative filtering for personalization, and LLM-based tracklist refinement in a retrieval-augmented generation workflow, deployed at scale on Deezer. Pilot deployment showed strong feasibility and positive user engagement, with 45% of generated playlists being listened to in subsequent days versus 27% for manually created playlists, and insights into commonly requested moods. The study demonstrates practical pathways for scalable, text-driven playlist creation and outlines future enhancements such as broader tag coverage, lyric-based enrichment, and conversational capabilities to further assist user discovery.

Abstract

The streaming service Deezer heavily relies on the search to help users navigate through its extensive music catalog. Nonetheless, it is primarily designed to find specific items and does not lead directly to a smooth listening experience. We present Text2Playlist, a stand-alone tool that addresses these limitations. Text2Playlist leverages generative AI, music information retrieval and recommendation systems to generate query-specific and personalized playlists, successfully deployed at scale.

Text2Playlist: Generating Personalized Playlists from Text on Deezer

TL;DR

This work tackles the gap between narrow, item-specific search and broad, exploratory text requests on Deezer by introducing Text2Playlist, a stand-alone text-to-playlist generator. The approach combines LLM-based tag extraction, tag-driven track retrieval, collaborative filtering for personalization, and LLM-based tracklist refinement in a retrieval-augmented generation workflow, deployed at scale on Deezer. Pilot deployment showed strong feasibility and positive user engagement, with 45% of generated playlists being listened to in subsequent days versus 27% for manually created playlists, and insights into commonly requested moods. The study demonstrates practical pathways for scalable, text-driven playlist creation and outlines future enhancements such as broader tag coverage, lyric-based enrichment, and conversational capabilities to further assist user discovery.

Abstract

The streaming service Deezer heavily relies on the search to help users navigate through its extensive music catalog. Nonetheless, it is primarily designed to find specific items and does not lead directly to a smooth listening experience. We present Text2Playlist, a stand-alone tool that addresses these limitations. Text2Playlist leverages generative AI, music information retrieval and recommendation systems to generate query-specific and personalized playlists, successfully deployed at scale.
Paper Structure (10 sections, 2 figures)

This paper contains 10 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Where to Explore and Save Music on Deezer
  3. Text2Playlist, a Playlist Generation Tool from Text
  4. Tags Extraction using LLM and Tracks Retrieval
  5. Leveraging Collaborative Filtering for Personalization
  6. Tracklist Refinement using LLM
  7. Deploying Text2Playlist on Deezer
  8. System Deployment
  9. Empirical Analysis
  10. Conclusion

Figures (2)

  • Figure 1: Interface of the "Text2Playlist" tool on the website version of Deezer, creating a personalized playlist generated from an idea given by a text from a Deezer user.
  • Figure 2: Overview of the Text2Playlist framework from Section \ref{['text2playlistsection']}, available for online requests in our production environment on Deezer.