Add ABA routing number recognizer (#348)

* aba routing number with checksum digit

* Update test_aba_routing_recognizer.py

Co-authored-by: Omri Mendels <omri374@users.noreply.github.com>

presidio-analyzer/presidio_analyzer/predefined_recognizers/__init__.py

@@ -1,3 +1,4 @@
+from .aba_routing_recognizer import AbaRoutingRecognizer
 from .credit_card_recognizer import CreditCardRecognizer
 from .crypto_recognizer import CryptoRecognizer
 from .domain_recognizer import DomainRecognizer
@@ -18,19 +19,20 @@
 NLP_RECOGNIZERS = {"spacy": SpacyRecognizer, "stanza": StanzaRecognizer}
 
 __all__ = [
+    "AbaRoutingRecognizer",
     "CreditCardRecognizer",
     "CryptoRecognizer",
     "DomainRecognizer",
     "EmailRecognizer",
     "IbanRecognizer",
     "IpRecognizer",
+    "NhsRecognizer",
     "SgFinRecognizer",
     "SpacyRecognizer",
     "StanzaRecognizer",
-    "NhsRecognizer",
     "UsBankRecognizer",
-    "UsLicenseRecognizer",
     "UsItinRecognizer",
+    "UsLicenseRecognizer",
     "UsPassportRecognizer",
     "UsPhoneRecognizer",
     "UsSsnRecognizer",

presidio-analyzer/presidio_analyzer/predefined_recognizers/aba_routing_recognizer.py

@@ -0,0 +1,60 @@
+from presidio_analyzer import Pattern, PatternRecognizer
+
+
+class AbaRoutingRecognizer(PatternRecognizer):
+    """
+    Recognizes American Banking Association (ABA) routing number.
+    Also known as routing transit number (RTN) and used to identify financial
+    institutions and process transactions.
+    """
+
+    PATTERNS = [
+        Pattern("ABA routing number (weak)", r"\b[0123678]\d{8}\b", 0.05,),
+        Pattern("ABA routing number",
+                r"\b[0123678]\d{3}-\d{4}-\d\b",
+                0.3,),
+    ]
+
+    CONTEXT = [
+        "aba",
+        "routing",
+        "abarouting",
+        "association",
+        "bankrouting",
+    ]
+
+    def __init__(
+        self,
+        patterns=None,
+        context=None,
+        supported_language="en",
+        supported_entity="ABA_ROUTING_NUMBER",
+        replacement_pairs=None,
+    ):
+        self.replacement_pairs = replacement_pairs or [("-", "")]
+        patterns = patterns if patterns else self.PATTERNS
+        context = context if context else self.CONTEXT
+        super().__init__(
+            supported_entity=supported_entity,
+            patterns=patterns,
+            context=context,
+            supported_language=supported_language,
+        )
+
+    def validate_result(self, pattern_text):
+        sanitized_value = self.__sanitize_value(pattern_text,
+                                                self.replacement_pairs)
+        return self.__checksum(sanitized_value)
+
+    @staticmethod
+    def __checksum(sanitized_value):
+        s = 0
+        for idx, m in enumerate([3, 7, 1, 3, 7, 1, 3, 7, 1]):
+            s += int(sanitized_value[idx]) * m
+        return s % 10 == 0
+
+    @staticmethod
+    def __sanitize_value(text, replacement_pairs):
+        for search_string, replacement_string in replacement_pairs:
+            text = text.replace(search_string, replacement_string)
+        return text

presidio-analyzer/tests/data/context_sentences_tests.txt

@@ -60,6 +60,9 @@ my DRIVER LICENSE is: 7774567901234
 US_DRIVER_LICENSE
 my DrIvEr LiCeNsE is: 7774567901234
 
+ABA_ROUTING_NUMBER
+routing number is: 101205681
+
 US_BANK_NUMBER
 my bank account number is 912803456
 
@@ -90,4 +93,4 @@ Special NRIC numbers e.g. S0000001I that are numerically significant have been i
 
 # Verify SG NRIC/FIN mixed case (e.g. lower case )
 FIN
-my fin is g3300299L
+my fin is g3300299L

presidio-analyzer/tests/test_aba_routing_recognizer.py

@@ -0,0 +1,38 @@
+import pytest
+
+from tests import assert_result
+from presidio_analyzer.predefined_recognizers import AbaRoutingRecognizer
+
+
+@pytest.fixture(scope="module")
+def recognizer():
+    return AbaRoutingRecognizer()
+
+
+@pytest.fixture(scope="module")
+def entities():
+    return ["ABA_ROUTING_NUMBER"]
+
+
+@pytest.mark.parametrize(
+    "text, expected_len, expected_positions, expected_score",
+    [
+        # Bank of America
+        ("121000358", 1, ((0, 9),), 1.0),
+        # Chase
+        ("3222-7162-7", 1, ((0, 11),), 1.0),
+        # Wells Fargo
+        ("121042882", 1, ((0, 9),), 1.0),
+        ("0711-0130-7", 1, ((0, 11),), 1.0),
+        # invalid ABA numbers
+        ("421042111", 0, (), -1.0),
+        ("1234-0000-0", 0, (), -1.0),
+    ],
+)
+def test_aba_routing_numbers(
+    text, expected_len, expected_positions, expected_score, recognizer, entities
+):
+    results = recognizer.analyze(text, entities)
+    assert len(results) == expected_len
+    for res, (st_pos, fn_pos) in zip(results, expected_positions):
+        assert_result(res, entities[0], st_pos, fn_pos, expected_score)

presidio-analyzer/tests/test_context_support.py

@@ -3,6 +3,7 @@
 
 from presidio_analyzer import PatternRecognizer, Pattern
 from presidio_analyzer.predefined_recognizers import (
+    AbaRoutingRecognizer,
     # CreditCardRecognizer,
     UsPhoneRecognizer,
     # DomainRecognizer,
@@ -23,6 +24,7 @@ def recognizers():
         "IP_ADDRESS": IpRecognizer(),
         "US_SSN": UsSsnRecognizer(),
         "PHONE_NUMBER": UsPhoneRecognizer(),
+        "ABA_ROUTING_NUMBER": AbaRoutingRecognizer(),
         "US_ITIN": UsItinRecognizer(),
         "US_DRIVER_LICENSE": UsLicenseRecognizer(),
         "US_BANK_NUMBER": UsBankRecognizer(),
@@ -58,9 +60,9 @@ def dataset(recognizers):
             raise ValueError(f"bad entity type {entity_type}")
 
         test_items.append((item, recognizer, [entity_type]))
-    # Currently we have 27 sentences, this is a sanity check
-    if not len(test_items) == 27:
-        raise ValueError(f"expected 27 context sentences but found {len(test_items)}")
+    # Currently we have 28 sentences, this is a sanity check
+    if not len(test_items) == 28:
+        raise ValueError(f"expected 28 context sentences but found {len(test_items)}")
 
     yield test_items
 
@@ -79,7 +81,10 @@ def test_text_with_context_improves_score(dataset, nlp_engine, mock_nlp_artifact
 
         assert len(results_without_context) == len(results_with_context)
         for res_wo, res_w in zip(results_without_context, results_with_context):
-            assert res_wo.score < res_w.score
+            if res_wo.score != 1.0:
+                assert res_wo.score < res_w.score
+            else:
+                assert res_wo.score <= res_w.score
 
 
 def test_context_custom_recognizer(nlp_engine, mock_nlp_artifacts):