This practical lesson puts Lesson 3.3 into practice. You run the Entity Consistency Checker on your own entity export, make a decision on each important cluster, and leave with a canonical naming table and your Consistency score.
What this lesson covers
- The Consistency Checker step by step, with screenshots
- The cluster filters and what each kind of cluster means
- A decision rule for merging, keeping apart or ignoring
- The two exports and where each goes
- Building and owning the mapping table
- Scoring Consistency
The Consistency Checker runs entirely in the browser and makes no API calls. Loaded with the demo site’s frozen export, it reads 102,752 mentions into 15,080 distinct forms, 759 of them carrying a Knowledge Graph identifier, and groups the variants into clusters with a suggested canonical form, the variants found, a kind, a severity, and the number of pages affected. The lesson shows each screen with what you should see, and explains the filters: same-MID variants, over-link review, containment review and punctuation.
The core of the practical is deciding. For each high-severity cluster you choose to merge the variants under one name, keep them apart because they are genuinely different things, or ignore them because they do not matter to the business. Those decisions go into the canonical naming table, exported from the tool and finished by hand with accepted variants and an owner, and the clusters file goes into the Scorecard for the Consistency score. The lesson also shows where over-linking candidates surface, ready for Lesson 3.4.
Because the tool only proposes candidates, the lesson is explicit about the limits of automation: nothing is merged automatically, case-only differences are switched off by default because they are almost never worth acting on, and the final say on every canonical name belongs to whoever owns the content.
This practical continues Module 3, Knowledge Graph Auditing & Entity Mapping, in The AI-powered Entity Audit, and leads into Lesson 3.4 on entity over-linking. It suits learners who want a written procedure beside the tool while they work through their own clusters, and content and editorial teams that need a naming standard writers will actually follow.