Skip to main content
Back to NLP Topics Reference for Knowledge Graph (KG) concepts. Applies to both Few-Shot and Ontology KG models unless noted. See Knowledge AI Introduction and Build a Knowledge Graph.
The KG is limited to 20,000 nodes/terms and 50,000 FAQs.

Terms / Nodes

Terms (nodes) are the structural building blocks of an ontology. They define domain concepts and are organized hierarchically.

Node Hierarchy

The KG engine doesn’t use parent-child relationships when evaluating FAQ matches. All nodes are considered equally, regardless of position.

Node / Term Types


Tags

Tags work like terms but aren’t shown in the KG ontology view (avoids clutter). You can add synonyms and traits to tags, just like terms.

Synonyms

Synonyms expand the set of words that can match a term, reducing the need for training with alternative questions.

Synonym Types


Traits

A trait is a collection of user utterances that characterize the nature of a query. Traits are applied to terms across the ontology to filter irrelevant suggestions.
  • If the user’s utterance matches a trait, the assistant searches only the nodes associated with that trait.
  • Nodes without the matching trait are ignored.
See Traits.

Intents

Each FAQ node can contain two types of intents:
  • For Few-Shot KGs, you can add all FAQs directly to the root node (no ontology required).
  • Default terms aren’t available in Few-Shot KGs, except when migrating from an Ontology KG (existing Default terms remain until edited, then become Organizer terms).
  • Path-Level Synonyms are supported only for Mandatory Terms and Tags.