Definition
A retrieval structure where documents are represented as high-dimensional, continuous vectors (embeddings) in a latent space, enabling semantic search through distance metrics like cosine similarity. Unlike keyword-based indices, it captures the contextual meaning and relationships between data points, allowing RAG systems to retrieve relevant context even when query terms do not exactly match the source text.
Search by semantic meaning and concepts rather than exact keyword or B-tree lookups.
"A 3D star map where concepts with similar 'vibes' are physically clustered together, regardless of their names."
- Vector Embedding(Prerequisite)
- Approximate Nearest Neighbor (ANN)(Component)
- Sparse Index(Complementary Retrieval Method)
- Cosine Similarity(Distance Metric)
Conceptual Overview
A retrieval structure where documents are represented as high-dimensional, continuous vectors (embeddings) in a latent space, enabling semantic search through distance metrics like cosine similarity. Unlike keyword-based indices, it captures the contextual meaning and relationships between data points, allowing RAG systems to retrieve relevant context even when query terms do not exactly match the source text.
Disambiguation
Search by semantic meaning and concepts rather than exact keyword or B-tree lookups.
Visual Analog
A 3D star map where concepts with similar 'vibes' are physically clustered together, regardless of their names.