WEBVTT 00:02.360 --> 00:12.480 Procedure Analyzer is a Python based service for detecting PII entities in text during analysis. 00:12.520 --> 00:23.880 It runs a set of different PII recognizers, each one in charge of detecting one or more PII entities 00:24.480 --> 00:26.400 using different mechanisms. 00:27.160 --> 00:38.000 It comes with a set of predefined recognizers, but it can be easily extended with other custom recognizers 00:38.000 --> 00:38.800 as well. 00:40.400 --> 00:50.880 Predefined and custom recognizers can leverage regex named entity recognition, also known as NER, 00:51.000 --> 00:56.040 and other types of logic, to detect PII in unstructured data. 00:58.320 --> 01:03.120 So here is the pip install Presidio analyzer. 01:03.680 --> 01:16.880 And here is the spacey Encore Web LG as the English language processing model for detecting the APIs. 01:17.160 --> 01:17.600 Right? 01:17.640 --> 01:20.440 Let's go ahead and install and download them. 01:22.800 --> 01:31.800 The following code sample will be responsible for setting up the analyzer engine, which would load 01:31.800 --> 01:41.200 the NLP model by default, and other PII recognizers which goes in here. 01:41.840 --> 01:42.280 Right. 01:42.440 --> 01:49.040 And then we'll call the analyzer to analyze the results for the phone number. 01:49.920 --> 01:59.760 In this case we are setting up the text to anonymize as the text that we want to analyze. 02:00.160 --> 02:04.920 And the entity that we want to detect is the phone number. 02:05.240 --> 02:05.760 Right. 02:05.800 --> 02:07.880 So here is the code for it. 02:07.880 --> 02:09.640 Let's go ahead and run this. 02:11.280 --> 02:14.800 Um, I missed the part where I had to import the, uh. 02:17.200 --> 02:18.200 The class here. 02:18.320 --> 02:19.440 Analyzer engine. 02:19.720 --> 02:20.160 Right. 02:20.480 --> 02:29.480 Uh, which also will in future require the pattern recognizer, which is a custom pattern that we will 02:29.480 --> 02:32.720 create for detecting our custom entities. 02:32.880 --> 02:33.160 Right. 02:33.200 --> 02:35.520 And now I'll go ahead and run this. 02:41.400 --> 02:41.920 Right. 02:42.040 --> 02:52.640 So now if you notice here it did detect phone number which starts from the um, uh, 46 as the count 02:52.680 --> 02:54.600 here until 58. 02:55.040 --> 02:55.560 Right. 02:55.600 --> 03:05.430 And the score is 0.75, which means that the, uh, the confidence level for, Saying that this particular 03:05.710 --> 03:09.390 entity is a phone number is very high, right? 03:09.630 --> 03:21.670 So that is how we use the analyzer engine to to analyze the text and uh, find the entity phone number. 03:22.110 --> 03:22.510 Right. 03:22.830 --> 03:35.350 Now let's go ahead and understand how the, the entire uh, class diagram for Procedure Analyzer is, 03:35.790 --> 03:36.590 is set up. 03:38.030 --> 03:44.910 So here is the, um, the class diagram for Procedure Analyzer. 03:46.310 --> 03:55.550 Here is the analyzer engine that we just went through to understand the, uh, understand and detect 03:55.590 --> 03:57.310 the, uh, the phone number. 03:57.590 --> 04:01.830 Uh, but if you notice here it has other components as well. 04:01.830 --> 04:05.550 In the, in the, in the hierarchy. 04:05.950 --> 04:06.150 Right. 04:06.150 --> 04:10.870 So here on the left hand side we have entity recognizer. 04:11.910 --> 04:18.110 Entity recognizer is an object that is responsible for detecting entities in text. 04:18.870 --> 04:24.710 It can be rule based recognizer machine learning model or a combination of both. 04:25.390 --> 04:25.790 Right. 04:26.070 --> 04:32.270 Then we have here uh, the recognizer registry. 04:33.150 --> 04:40.310 Uh, it is a registry that contains all the entity recognizers that are available in the Presidio. 04:40.910 --> 04:48.190 The analyzer engine uses the registry to detect entities in the text. 04:49.390 --> 04:49.790 Right. 04:50.430 --> 05:00.670 Then here we have the NLP engine, which is an object that holds the NLP model that is used by the analyzer 05:00.990 --> 05:01.550 engine. 05:05.030 --> 05:16.190 You can use a variety of different NLP engines to, uh, to detect the, uh, the entities in the text. 05:16.230 --> 05:28.510 You can use Spacy, transformer or Stanza NLP engines to, uh, to run the detection uh, in the text. 05:29.270 --> 05:31.710 And then here we have context aware enhancer. 05:32.270 --> 05:37.790 It's a module that enhances the detection of entities by using context of the text. 05:39.830 --> 05:44.270 One particular class that is not covered here is the pattern recognizer. 05:44.830 --> 05:53.110 Uh, it is a type of entity recognizer that uses regular expressions to detect entities in the text. 05:53.270 --> 05:57.670 Um, we'll cover that as part of another video with hands on activity. 05:58.110 --> 06:04.790 But here is the, uh, high level understanding of how Presidio Analyzer works. 06:05.630 --> 06:06.230 Thank you. 06:06.310 --> 06:07.190 I'll see you in the next.