fix(analyzer): add Korean context terms to KrRrnRecognizer (#2212)

presidio-analyzer/presidio_analyzer/predefined_recognizers/country_specific/korea/kr_rrn_recognizer.py

@@ -43,6 +43,12 @@ class KrRrnRecognizer(PatternRecognizer):
     ]
 
     CONTEXT = [
+        "주민등록번호",
+        "주민번호",
+        "주민등록증",
+        "주민등록",
+        "신분증",
+        "본인인증",
         "Korean RRN",
         "Korean Resident Registration Number",
         "Resident Registration Number",

presidio-analyzer/tests/test_kr_rrn_recognizer.py

@@ -1,16 +1,21 @@
 import pytest
+from presidio_analyzer.predefined_recognizers import KrRrnRecognizer
 
 from tests import assert_result_within_score_range
-from presidio_analyzer.predefined_recognizers import KrRrnRecognizer
+
 
 @pytest.fixture(scope="module")
 def recognizer():
+    """Return KrRrnRecognizer instance."""
     return KrRrnRecognizer()
 
+
 @pytest.fixture(scope="module")
 def entities():
+    """Return supported entity list."""
     return ["KR_RRN"]
 
+
 @pytest.mark.parametrize(
     "text, expected_len, expected_positions, expected_score_ranges",
     [
@@ -20,15 +25,19 @@ def entities():
         ("000505-3637892", 1, ((0, 14),), ((0.5, 0.5),), ),
         ("0005053637892", 1, ((0, 13),), ((0.5, 0.5),), ),
         ("His Korean RRN is 960121-1234567", 1, ((18, 32),), ((0.5, 0.5),), ),
-
+        ("주민등록번호: 960121-1234567", 1, ((8, 22),), ((0.5, 0.5),), ),
+        ("신분증 확인: 960121-1234567", 1, ((8, 22),), ((0.5, 0.5),), ),
+        ("본인인증 960121-1234567", 1, ((5, 19),), ((0.5, 0.5),), ),
+
         # Valid RRNs, strong match by validate_result()
         ("960121-1021413", 1, ((0, 14),), ((1.0, 1.0),), ),
         ("9601211021413", 1, ((0, 13),), ((1.0, 1.0),), ),
         ("050912-2000019", 1, ((0, 14),), ((1.0, 1.0),), ),
         ("0509122000019", 1, ((0, 13),), ((1.0, 1.0),), ),
         ("His RRN is 9601211021413", 1, ((11, 24),), ((1.0, 1.0),), ),
-
-        # Invalid RRNs 
+        ("주민번호는 9601211021413입니다", 1, ((6, 19),), ((1.0, 1.0),), ),
+
+        # Invalid RRNs
         ("001332-1234567", 0, (), (),),
         ("0013321234567", 0, (), (),),
         ("960121+1021413", 0, (), (),),
@@ -48,6 +57,7 @@ def test_when_all_rrns_then_succeed(
     entities,
     max_score,
 ):
+    """Test recognition and validation of Korean RRN across valid and invalid inputs."""
     results = recognizer.analyze(text, entities)
     assert len(results) == expected_len
     for res, (st_pos, fn_pos), (st_score, fn_score) in zip(
@@ -58,3 +68,90 @@ def test_when_all_rrns_then_succeed(
         assert_result_within_score_range(
             res, entities[0], st_pos, fn_pos, st_score, fn_score
         )
+
+
+def test_korean_context_keywords_in_rrn_recognizer(recognizer):
+    """Test that Korean context keywords are present in KrRrnRecognizer.context."""
+    expected_korean_keywords = [
+        "주민등록번호",
+        "주민번호",
+        "주민등록증",
+        "주민등록",
+        "신분증",
+        "본인인증",
+    ]
+    for keyword in expected_korean_keywords:
+        assert keyword in recognizer.context
+
+
+@pytest.mark.parametrize(
+    "context_word",
+    ["주민등록번호", "주민번호", "주민등록증", "주민등록", "신분증", "본인인증"],
+)
+def test_when_korean_context_used_then_score_enhanced(recognizer, context_word):
+    """Test that Korean context keywords enhance confidence score from 0.5 to 0.85."""
+    from presidio_analyzer.context_aware_enhancers import LemmaContextAwareEnhancer
+    from presidio_analyzer.nlp_engine import NlpArtifacts
+
+    text = f"{context_word} 960121-1234567"
+    results = recognizer.analyze(text, ["KR_RRN"])
+    assert len(results) == 1
+    assert results[0].score == 0.5
+
+    tokens = [context_word, "960121-1234567"]
+    tokens_indices = [0, len(context_word) + 1]
+    artifacts = NlpArtifacts(
+        entities=[],
+        tokens=tokens,
+        tokens_indices=tokens_indices,
+        lemmas=tokens,
+        nlp_engine=None,
+        language="ko",
+    )
+    artifacts.keywords = [context_word, "960121-1234567"]
+
+    enhancer = LemmaContextAwareEnhancer()
+    enhanced_results = enhancer.enhance_using_context(
+        text, results, artifacts, [recognizer]
+    )
+
+    assert len(enhanced_results) == 1
+    assert enhanced_results[0].score == 0.85
+    assert (
+        enhanced_results[0].analysis_explanation.supportive_context_word
+        == context_word
+    )
+
+
+def test_when_no_korean_context_then_score_not_enhanced(recognizer):
+    """Test that unrelated context does not enhance confidence score."""
+    from presidio_analyzer.context_aware_enhancers import LemmaContextAwareEnhancer
+    from presidio_analyzer.nlp_engine import NlpArtifacts
+
+    text = "random text 960121-1234567"
+    results = recognizer.analyze(text, ["KR_RRN"])
+    assert len(results) == 1
+    assert results[0].score == 0.5
+
+    tokens = ["random", "text", "960121-1234567"]
+    tokens_indices = [0, 7, 12]
+    artifacts = NlpArtifacts(
+        entities=[],
+        tokens=tokens,
+        tokens_indices=tokens_indices,
+        lemmas=tokens,
+        nlp_engine=None,
+        language="ko",
+    )
+    artifacts.keywords = ["random", "text", "960121-1234567"]
+
+    enhancer = LemmaContextAwareEnhancer()
+    enhanced_results = enhancer.enhance_using_context(
+        text, results, artifacts, [recognizer]
+    )
+
+    assert len(enhanced_results) == 1
+    assert enhanced_results[0].score == 0.5
+    assert (
+        enhanced_results[0].analysis_explanation.supportive_context_word == ""
+    )