Knowledge graph
A knowledge graph stores things as entities and relationships as triples. This makes connections explicit instead of leaving them only inside document text.
Why use a graph
Section titled “Why use a graph”Text search finds documents that contain a term. A graph query can follow relationships across records.
For example, separate documents can contribute:
(Tesla, has_ceo, Elon Musk)(Tesla, headquartered_in, Austin)
The graph can return both facts from the same Tesla entity even when the source documents are different.
What yggy stores
Section titled “What yggy stores”- Entities have a label, type, and optional aliases and description.
- Triples connect a subject entity to an object entity with a predicate.
- Literal properties attach typed values to an entity.
- Documents provide source context for extracted triples.
- Projects scope graph data and access.
Triples can include confidence, source text, validity dates, and other metadata. These fields are optional.
Source and review
Section titled “Source and review”An extracted triple keeps a document relationship or source identifier. Where source text is available, the Graph page can show the supporting excerpt.
Entity resolution and contradiction detection can produce Review items. Confident same-name duplicates merge automatically unless the project asks to review every case. A human still decides unclear merges and which triple to keep.