GraphRAG Glossary

Chunk — a common name for a part of a larger text, resulting from text splitting
- Chunking — the technique of splitting a text into smaller parts
Cypher — a common name for ISO-GQL (like Javascript to Ecmascript)

Generative AI (GenAI) — artificial intelligence utilizing foundation models to generate content
Graph-based indexing —
- See Graph Retrieval-Augmented Generation: A Survey
Graph embedding — a vector embedding of a node, relationship, or other graph structure
Graph-Guided Retrieval —
- See Graph Retrieval-Augmented Generation: A Survey
GraphRAG — RAG using a knowledge graph

Labeled Property Graph (LPG) — an alternative name for the Property Graph Model
LLM — Large Language Model, a GenAI foundation model specializing in text
Node — generally, a thing somehow connected to other things
1. in a property graph, a node is a data record connected by relationships
2. in mathematics, a point where two or more curves, lines, or edges meet (a vertex)
Property Graph Model (PGM) — the data model of a graph database composed of nodes and relationships
RAG — Retrieval Augmented Generation is a software design pattern for integrating a data source into the orchestration layer of a GenAI application
Relationship — a data record containing an ordered pair of nodes in the Property Graph Model
Text embedding — a vector embedding for text, intended to place the text within an information space based on semantics. For example, ""

Vector — a matrix with one row or column
Vector embedding — a vector representation of other data. Can be derived from words, images or even graph structures
Vector similarity — the “closeness” of two vectors within a vector space, implying that the vectors are similar according to the embedding approach
Vertex —
1. in geometry, the meeting point of two lines that form an angle