Textual User Taste: Natural-Language User Context for Foundation-Model Recommender System at Scale

2026-09-28

AI Summary

This study addresses the opacity of behavioral embeddings in foundation model-based recommender systems and their incompatibility with natural language interaction by proposing structured taste profiles as interpretable interfaces. Methodologically, an end-to-end production pipeline is constructed using large language model prompt engineering and compression techniques, enabling user steering and deep integration with downstream systems. A multi-dimensional evaluation framework is further introduced to accommodate the absence of unique ground truth. Experimental results demonstrate that this approach improves track prediction MRR by 0.6% and search ranking NDCG@7 by 2.2%, effectively achieving interpretability and positive guidance for large-scale user contexts while delineating its application boundaries.

Abstract

Foundation model recommender systems require user context that can be consumed by large language models, reasoned over, and refined through natural-language interaction. Traditional behavioral embedding vectors remain highly effective for retrieval and ranking, but they are opaque to users and not natively expressed for language model workflows. We present Textual User Taste, a system that generates structured natural-language taste profiles from listening behavior, interaction signals, content metadata, and optional user feedback, and deploys them to millions of Spotify users. We describe the end-to-end production lifecycle required to generate, evaluate, optimize, and maintain these representations at industrial scale, including prompt development and compression, user steering, and integration with downstream personalization systems. Because no unique ground-truth taste profile exists, we introduce a multi-faceted evaluation framework to evaluate taste profiles as a production representation: they carry user-specific predictive signal independently, and when integrated with behavioral embeddings, improve MRR by 0.6% for future-track prediction and NDCG@7 by 2.2% for search ranking. Our evaluation also reveals that taste profiles support positive natural-language steering, while exposing important limitations, including challenges with negation and short-term temporal adaptation. These findings position taste profiles not as replacements for behavioral embeddings, but as an interpretable and steerable interface between evolving user context and foundation-model recommender systems.

Problem

• Recommender System
• Foundation Model
• User Context
• Behavioral Embeddings
• Taste Profile

Innovation

• Textual User Taste
• Foundation-Model Recommender System
• Natural-Language Steering
• Multi-faceted Evaluation Framework
• Behavioral Embeddings