中央研究院 資訊科學研究所

Query expansion is a powerful technique that transforms basic user queries into richer, more comprehensive search terms for Retrieval-Augmented Generation (RAG) systems. When users ask questions in natural language, they often use different terminology than what appears in the knowledge base documents. By automatically expanding queries with synonyms, related concepts, and contextual terms, RAG systems can bridge this vocabulary gap and retrieve truly relevant information that might otherwise be missed. This presentation demonstrates how implementing query expansion can significantly enhance the accuracy and helpfulness of LLM-based systems, allowing them to better understand user intent and deliver more precise answers from your knowledge bases.