KGConstructionOptions
Options & Configuration · AiDotNet.RetrievalAugmentedGeneration.Graph.Construction
Configuration options for automated knowledge graph construction from text.
For Beginners
These settings control how the KG construction pipeline works:
- MaxChunkSize: How many characters per text chunk (smaller = more precise entity extraction)
- ChunkOverlap: Overlap between chunks to avoid missing entities at boundaries
- EntityConfidenceThreshold: Minimum confidence to accept an extracted entity
- EnableEntityResolution: Whether to merge similar entity names (e.g., “Einstein” and “A. Einstein”)
- EntitySimilarityThreshold: How similar two entity names must be to merge them
Constructors
| Constructor | Summary |
|---|---|
KGConstructionOptions() |
Properties
| Property | Summary |
|---|---|
bool? EnableEntityResolution { get; set; } |
Whether to merge similar entity names via string similarity. |
bool? ExtractClaims { get; set; } |
Whether to additionally ask the LLM to extract claims/covariates (Microsoft GraphRAG style) about entities. |
bool? UseLlmExtraction { get; set; } |
Whether to use the injected LLM text generator for entity/relation extraction when one is available. |
double? EntityConfidenceThreshold { get; set; } |
Minimum confidence threshold to accept an extracted entity. |
double? EntitySimilarityThreshold { get; set; } |
Minimum string similarity (0-1) for merging entity names. |
int? ChunkOverlap { get; set; } |
Character overlap between adjacent chunks. |
int? MaxChunkSize { get; set; } |
Maximum characters per text chunk for entity extraction. |
int? MaxEntitiesPerSentence { get; set; } |
Maximum number of entities per sentence for co-occurrence relation generation. |