Computing system that is configured to identify search results based upon multi-modality searches
Topics: Chunk Relevance, LLM Readability, LLMO / GEO, Microsoft, Passage based retrieval, Probably in use, Query Fan Out, Retrieval Augmented Generation (RAG), Semantic Search
This Microsoft patent describes a computing system designed to improve how we search for files that contain different types of content, such as text and images. Instead of relying solely on keywords, the system creates a “semantic embedding” of a user’s natural language query to understand its deeper meaning. It then performs multiple parallel searches across different specialized indices—using both keywords and semantic meanings—to deliver highly relevant results that traditional search methods often miss.
