How is confidenceLevel calculated?

String similarity between the submitted name and the returned entity name. Exact or nearly-exact matches approach 1.0; more differences lower the score.

‍