0

I am trying to filter results by minimum precision to exclude any PII names that are too low in confidence (.5) using python and Azures PII cognitive service.

response=client.recognize_pii_entities(documents,language="en",categories_filter=['Person'], minimumPrecision=[.8])

the request() gets an unexpected key word argument for "minimumPrecision"

categories_filter was different from the piiCategories that the azure documentation says to use and that is the functional parameter so I am wondering if minimumPrecision is also actually supposed to be different, but I cannot find it in the docs.

William Rogers
  • 3
  • 1

1 Answers1

0

The client library does not support a minimumPrecision parameter. The REST API does not either. Can you share the Azure documentation which shows usage of this parameter?

You should be able to filter the entity results on the confidence_score for the same effect:

response = client.recognize_pii_entities(documents, language="en", categories_filter=['Person'])
results = [entity for entity in res.entities for res in result if entity.confidence_score > 0.8]
Krista
  • 261
  • 1
  • 5
  • that will have to do then, thank you. here is the link tot he docs https://learn.microsoft.com/en-us/azure/search/cognitive-search-skill-pii-detection – William Rogers May 11 '22 at 20:51
  • Ah I see. It looks like this documentation is showing PII detection as a Cognitive Search skill. Azure Cognitive Search is accessed through a different client library. You could look at the azure-search-documents package if you want support for minimumPrecision. Although since Search is using the Language service anyway, you'll get the same results as you have now with above code. – Krista May 11 '22 at 21:09