Select Taxonomy Type
Select Source Collection
Select Enrichment Fields
Select Retriever
Configure Input Fields
Content & Topic Tagging
Product Attribute Enrichment
Compliance & Safety Classification
Example: Content Tagging
Multi-Level Data Organization
Linking Disparate Datasets
face_collection
) to the scenes they appear in (scene_collection
), or connect product mentions in customer reviews (review_collection
) to entries in your official product catalog (product_collection
). This enables powerful graph-like queries. Detailed examples typically involve pipeline configurations rather than just taxonomy creation.Explicit Hierarchy Enrichment & Retrieval
parent
links).parent
, depth
, or path
calculated during ingestion.
Example: A document classified under Photography
is automatically enriched with Visual
and Media
tags if those are its defined ancestors.Visual
) retrieves items tagged with its descendants (e.g., Photography
, Illustration
).WHERE category IN subtree('Visual')
).Implicit Relationship Enrichment & Retrieval
face_id
or embedding similarity).face_id
from a face detector feeds into a scene taxonomy retriever).face_id: 123
), and a subsequent step uses this ID to query a Scene taxonomy, enriching the original document with relevant scene information associated with face_id: 123
.
Feature | Explicit Hierarchy (Intra-Taxonomy) | Implicit Relationships (Inter-Taxonomy / Dynamic) |
---|---|---|
Defined by | Schema (parent , structure) | Data behavior, shared IDs, retriever chaining |
Enrichment | Rule-based, tree traversal | Dynamic, input-driven enrichment chains |
Retrieval | IN subtree("X") , static filters | Chained stages (e.g., Face → Scene → Cluster) |
Query Composition | Single query, expanded filters | Multi-step query pipeline, composite queries |
Analogy | Folder structure / Tree | SQL JOINs / Graph Traversal |
Examples | Category trees, org charts | Face → Scene → Episode links, Product Mentions |
Design Thoughtful Hierarchies
Optimize Retriever Performance
Consider Materialization Strategy
Test with Representative Data
face_id
found from a face query used as input into a scene_taxonomy
retriever)"John Wick"
→ character_id
)