Find taxonomy terms that are near-duplicates of each other (typos, plurals, synonyms) and offer to merge them in one place.
How it works
Term cleanup tackles the long-tail problem on every long-running content site: a tag taxonomy that contains AI, A.I., ai, Artificial Intelligence, and artificial-intelligence as five separate terms. ClassifAI generates an embedding for each top-level term in the taxonomies you select, runs a vector similarity search across them, and presents an admin screen of suspected duplicate clusters. Administrators can review each cluster and merge selected terms into a chosen primary term in a single action; merged terms behave like a normal WordPress term merge, so all posts assigned to the merged terms are re-assigned to the primary.
Because the comparison is embedding-based rather than string-based, the feature catches cases that a regex or fuzzy-string match would miss — for example, an entity that is spelled correctly in two different languages.
Configuration
- Which taxonomies are scanned for similar terms.
- Similarity threshold for proposed merge clusters.
- Provider and model selection.
- Allowed roles and an allowed-users list for granular access control.
Providers
- OpenAI Embeddings
- Azure OpenAI Embeddings
- Ollama Embeddings (locally hosted)
Caveats
- Requires ElasticPress as the vector index.
- Operates on top-level terms only — if a merged term has children they become top-level after the merge (standard WordPress behaviour).