Removing pre-existing recognizers

[katiejim]

Hello!

I am trying to remove some of the pre-defined recognizers as they are causing confusion with custom regex recognizers that are specific to my industry. I have tried using analyzer.registry.remove_recognizer(au_tfn_recognizer) to disable the recognizer for AU tax file numbers, but I get the error "NameError: name 'au_tfn_recognizer' is not defined"

The analyzer.registry.remove_recognizer() command seems to only work when disabling recognizers that I have made myself. Please can you let me know if I am missing something, or if it simply isn't possible to disable the pre-defined recognizers?

Thanks so much
Kate

[omri374]

Hi @katiejim,

The recognizer name is the class name, so to remove this recognizer, the right call would be:

analyzer.registry.remove_recognizer("AuTfnRecognizer")

Alternatively, you could start with an empty list of recognizers and add your own:

from presidio_analyzer import AnalyzerEngine, RecognizerRegistry

registry = RecognizerRegistry()
registry.add_recognizer(my_recognizer)
analyzer = AnalyzerEngine(registry=registry)
analyzer.analyze(...) # would only return values detected by my_recognizer

[katiejim]

@omri374 thank you so much, that's totally solved the problem.

[MarPeg53]

Hello everyone and thanks for the support. I have a similar question. When I remove "DateRecognizer" it does not work. What am I doing wrong?

titles_recognizer = PatternRecognizer(supported_entity="TITLE",
                                      deny_list=["Mr.","Mrs.","Miss"])
registry = RecognizerRegistry()
registry.load_predefined_recognizers()

registry.add_recognizer(titles_recognizer)
registry.remove_recognizer("DateRecognizer")

analyzer = AnalyzerEngine(registry=registry)

text="His name is Mr. Jones and he is born 3. July"
results = analyzer.analyze(text=text, language="en")
print(results)

[MarPeg53]

@omri374 thank you for your help! Can you tell me how to remove DATE_TIME? Probably not with registry.remove_recognizer("DATE_TIME").

[omri374]

Here's an example:

from presidio_analyzer import AnalyzerEngine

analyzer = AnalyzerEngine()
entities = analyzer.get_supported_entities()

entities_without_date = [ent for ent in entities if ent != "DATE_TIME"]

analyzer.analyze(text="They called me at 6pm", language = "en")
analyzer.analyze(text="They called me at 6pm", language = "en", entities = entities_without_date)

[akshatagawali]

Hi can you give answer please. actually I want to use only particular specified built-in recognizers in presidio not all how I can do that can you please explain.

[akshatagawali]

how can we remove particular built in recognizer from presidio. can we remove it like custom recognizers ??

[omri374]

Hi, one option is to define recognizers via yaml, and only keep those that you need. Another option is to remove those that you don't need via code (like mentioned in this thread). Third is to start with an empty list and adding those that you need. For that, you can instantiate RecognizerRegistry and not call load_predefined_recognizers. There are examples to this in the samples section and across different issues answered.