Agent-based searching over heterogeneous data sources
Topics: Data Mining, Graph RAG, Knowledge Graph, Microsoft, Probably in use, Retrieval Augmented Generation (RAG), Search Intent, Semantic Search
This Microsoft patent describes a system that allows users to search across many different types of data sources using plain natural language, without needing to know what databases exist or what query languages they use. It uses a combination of AI agents, large language models, and a “data map” that captures how different data sources relate to each other. The system figures out what the user actually wants, finds the right data sources, and sends specialized search agents to retrieve relevant information. The result is then compiled into a clear, coherent answer for the user.
