Query Attribute Modeling: Improving search relevance with Semantic Search and Meta Data Filtering

The exponential growth of e-commerce has created vast product catalogs where traditional keyword, semantic or hybrid search struggles to balance precision with relevance, frequently overlooking attribute constraints or misinterpreting user intent. This study introduces Query Attribute Modeling (QAM), a hybrid framework that enhances search precision and relevance through a two-step process. First, item descriptions and titles are decomposed into structured attributes and stored as metadata key-value pairs alongside item description embeddings. Second, QAM decomposes open text queries into structured metadata tags and semantic elements, enabling focused retrieval by automatically extracting metadata filters from free-form text queries and reducing noise. Experimental evaluation using the Amazon Toys Reviews dataset (10,000 unique items with 40,000+ reviews and detailed product attributes) demonstrated QAM's superior performance, achieving a mean average precision at 5 (mAP@5) of 52.99%. QAM showed substantial improvements: 28.67% over BM25 keyword-based search, 6.5% over semantic search, 8.58% over cross-encoder reranking, and 9.96% over hybrid search combining encoder embeddings and BM25 results using Reciprocal Rank Fusion. The results establish QAM as a robust solution for Enterprise Search applications, particularly in e-commerce systems.