Unifying Search and Recommendation: A Generative Paradigm Inspired by Information Theory

2025-04-09

AI Summary

This paper addresses the limitations of disjoint modeling in search-and-recommendation (S&R) systems—namely, inconsistent user/item representations, gradient conflicts in discriminative models, and heavy reliance on hand-crafted designs—by proposing GenSR, the first generative unified framework for S&R. Methodologically, GenSR introduces: (1) an information-theoretic generative paradigm for joint S&R modeling; (2) a task-prompt-driven parameter subspace partitioning mechanism to decouple task-specific representations; and (3) a theoretical interpretation of joint optimization grounded in mutual information maximization. The framework integrates dual representation learning (collaborative + semantic history), contrastive learning, and instruction tuning, with task prompts guiding generative decoding. Evaluated on two public benchmarks, GenSR achieves state-of-the-art performance, significantly improving both search relevance and recommendation accuracy—demonstrating the effectiveness and generalizability of the generative paradigm for unified S&R modeling.

Technology Category

• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Search
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Recommender systems and search engines serve as foundational elements of online platforms, with the former delivering information proactively and the latter enabling users to seek information actively. Unifying both tasks in a shared model is promising since it can enhance user modeling and item understanding. Previous approaches mainly follow a discriminative paradigm, utilizing shared encoders to process input features and task-specific heads to perform each task. However, this paradigm encounters two key challenges: gradient conflict and manual design complexity. From the information theory perspective, these challenges potentially both stem from the same issue -- low mutual information between the input features and task-specific outputs during the optimization process. To tackle these issues, we propose GenSR, a novel generative paradigm for unifying search and recommendation (S&R), which leverages task-specific prompts to partition the model's parameter space into subspaces, thereby enhancing mutual information. To construct effective subspaces for each task, GenSR first prepares informative representations for each subspace and then optimizes both subspaces in one unified model. Specifically, GenSR consists of two main modules: (1) Dual Representation Learning, which independently models collaborative and semantic historical information to derive expressive item representations; and (2) S&R Task Unifying, which utilizes contrastive learning together with instruction tuning to generate task-specific outputs effectively. Extensive experiments on two public datasets show GenSR outperforms state-of-the-art methods across S&R tasks. Our work introduces a new generative paradigm compared with previous discriminative methods and establishes its superiority from the mutual information perspective.

Problem

• Unifies search and recommendation tasks in one model
• Addresses gradient conflict and manual design complexity issues
• Enhances mutual information between input and task outputs

Innovation

• Generative paradigm unifying search and recommendation
• Task-specific prompts partition parameter space
• Dual representation learning and contrastive learning