Canonical Query

In the semantic audit clustering pipeline, a canonical query is determined after keywords are split into clusters: it's the phrase that best captures the cluster's topic and is closest to its embedding centroid. The canonical query becomes the seed query for further analysis — query expansion, knowledge graph construction, and content planning.

The difference between canonical query and root query: canonical is the cluster's 'name', root is the most basic form of the query (e.g., canonical 'how to calculate forced inheritance' → root 'forced inheritance').