Loading content
Entity-first retrieval first maps a question to known entities in the knowledge graph. From these entry points, relationships, neighboring entities, and associated text units are traversed and prioritized. The pattern is especially useful for questions about concrete actors, objects, and their multi-hop relationships.
Use this profile to understand the building block briefly, place it in the model, and open related building blocks.
Theoretical construct: explains a term, principle, or mental model.
What organizes, connects, or makes decisions possible.
Entity-first retrieval is a retrieval pattern that maps a question to entities in a knowledge graph first and expands the answer context from there.
The pattern sits in the GraphRAG and knowledge-graph retrieval space for questions anchored in concrete entities. Instead of starting from an arbitrary text chunk, the question is first mapped to known graph nodes; from there, relationships, neighboring entities, and matching text units are combined. This helps answer entity-centered questions that depend on multi-hop context.
Think of the search as moving from a pin to its surrounding neighborhood. The system first identifies candidate entities such as people, products, or places. From each starting node, it collects edges, neighboring nodes, and supporting evidence, then ranks and compresses them by relevance. The graph supplies structure; the text supplies justification.
A question is aligned to known nodes or references in the graph.
Chosen entry points determine the focus of the subsequent search.
Direct relationships and adjacent nodes provide the first layer of context.
Several edge traversals connect the starting entity to farther, but relevant, facts.
Linked text passages ground graph hits in original wording.
The semantic structure of entities and relationships that the search builds on.
This pattern is useful when questions target specific actors, objects, organizations, or processes and when relationships are part of the answer. It works well for research, analysis, and assistant systems once the knowledge graph is populated reliably. It is weaker when entity recognition is ambiguous, edges are missing, or the question is broad; in those cases, chunk-first retrieval is often more robust.
Where this building block is located in the topic model.
No structure path available.
Explore how this building block connects to concepts, methods, technologies, and tools.
These sources establish the term and its professional meaning.
All direct connections of the current building block in a compact text view.
This classification shows where the building block typically matters, how demanding it is, and what kind of impact it has in the model.
The level within the organization (enterprise, domain, team) at which the AssetBlock is applied.