CardRewriter: Leveraging Knowledge Cards for Long-Tail Query Rewriting on Short-Video Platforms

2025-10-11

AI Summary

Long-tail queries on short-video platforms frequently suffer from spelling errors, incomplete expressions, and ambiguous user intent, leading to suboptimal retrieval matching. Existing large language models (LLMs) lack domain-specific knowledge—such as short-video semantics, live-streaming dynamics, micro-drama narrative structures, and social relationships—limiting their query rewriting effectiveness. To address this, we propose a two-stage knowledge-card-based query rewriting framework: first, constructing heterogeneous knowledge cards integrating video semantics, host–audience relationships, and plot-level structures; second, leveraging domain-aware supervised fine-tuning and Groupwise Relative Policy Optimization (GRPO) to guide LLMs in generating high-fidelity rewrites. Our key innovations include a knowledge-card-driven domain-knowledge injection mechanism and a short-video-intent-oriented custom training paradigm. Offline experiments demonstrate significant gains in rewrite accuracy; online A/B tests show +3.2% long-watch rate, +2.7% CTR, and −18.5% active re-query rate. The system is deployed at Kuaishou, serving over 100 million users daily.

Technology Category

• Data Mining & Knowledge Management: Intelligent Query Processing
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Knowledge Engineering

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Short-video platforms have rapidly become a new generation of information retrieval systems, where users formulate queries to access desired videos. However, user queries, especially long-tail ones, often suffer from spelling errors, incomplete phrasing, and ambiguous intent, resulting in mismatches between user expectations and retrieved results. While large language models (LLMs) have shown success in long-tail query rewriting within e-commerce, they struggle on short-video platforms, where proprietary content such as short videos, live streams, micro dramas, and user social networks falls outside their training distribution. To address this challenge, we introduce extbf{CardRewriter}, an LLM-based framework that incorporates domain-specific knowledge to enhance long-tail query rewriting. For each query, our method aggregates multi-source knowledge relevant to the query and summarizes it into an informative and query-relevant knowledge card. This card then guides the LLM to better capture user intent and produce more effective query rewrites. We optimize CardRewriter using a two-stage training pipeline: supervised fine-tuning followed by group relative policy optimization, with a tailored reward system balancing query relevance and retrieval effectiveness. Offline experiments show that CardRewriter substantially improves rewriting quality for queries targeting proprietary content. Online A/B testing further confirms significant gains in long-view rate (LVR) and click-through rate (CTR), along with a notable reduction in initiative query reformulation rate (IQRR). Since September 2025, CardRewriter has been deployed on Kuaishou, one of China's largest short-video platforms, serving hundreds of millions of users daily.

Problem

• Addresses long-tail query issues like spelling errors and ambiguous intent on short-video platforms
• Enhances query rewriting using domain-specific knowledge cards to capture user intent
• Improves retrieval effectiveness for proprietary content including videos and live streams

Innovation

• LLM framework incorporates domain-specific knowledge cards
• Two-stage training optimizes query relevance and retrieval
• Knowledge aggregation enhances long-tail query rewriting accuracy