feat(analyzer): add Korean-language context terms to KrRrnRecognizer

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

@@ -49,6 +49,10 @@ class KrRrnRecognizer(PatternRecognizer):
         "RRN",
         "rrn",
         "rrn#",
+        "주민등록번호",
+        "주민번호",
+        "신분증",
+        "본인인증",
     ]
 
     def __init__(

presidio-analyzer/tests/test_kr_rrn_recognizer.py

@@ -1,7 +1,10 @@
 import pytest
+from presidio_analyzer.context_aware_enhancers import LemmaContextAwareEnhancer
+from presidio_analyzer.nlp_engine import NoOpNlpEngine
+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():
@@ -58,3 +61,59 @@ 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_when_korean_context_terms_then_present_in_context(recognizer):
+    """Korean terms must be in context so enhancement can fire on Korean text."""
+    for korean_term in ["주민등록번호", "주민번호", "신분증", "본인인증"]:
+        assert korean_term in recognizer.context
+
+
+@pytest.mark.parametrize(
+    "text, start, end",
+    [
+        # context term before the number: the default prefix window
+        ("주민등록번호 960121-1021413 확인 부탁드립니다", 7, 21),
+        # context term after the number: the suffix window, off by default
+        ("960121-1021413 주민번호로 본인인증 완료", 0, 14),
+    ],
+)
+def test_when_korean_context_on_either_side_then_pattern_match_is_kept(
+    text, start, end, recognizer, entities
+):
+    """Korean context text on either side must not disturb the match itself."""
+    results = recognizer.analyze(text, entities)
+    assert len(results) == 1
+    assert_result_within_score_range(
+        results[0], entities[0], start, end, 1.0, 1.0
+    )
+
+
+@pytest.mark.parametrize(
+    "korean_term", ["주민등록번호", "주민번호", "신분증", "본인인증"]
+)
+def test_when_korean_context_term_is_supplied_then_score_is_boosted(
+    korean_term, recognizer, entities
+):
+    """Each new Korean term must actually raise the score through the enhancer.
+
+    Context taken from the surrounding text needs a Korean NLP model, which
+    the test environment does not ship, so the boost is exercised through the
+    explicit ``context`` argument, the same way ``test_context_support`` does.
+    A checksum-invalid number is used so there is room above the base score.
+    """
+    text = "960121-1234567"
+    engine = NoOpNlpEngine(models=[{"lang_code": "ko", "model_name": "no_op"}])
+    engine.load()
+    nlp_artifacts = engine.process_text(text, "ko")
+    raw = recognizer.analyze(text, entities, nlp_artifacts)
+    enhancer = LemmaContextAwareEnhancer()
+
+    without = enhancer.enhance_using_context(text, raw, nlp_artifacts, [recognizer])
+    boosted = enhancer.enhance_using_context(
+        text, raw, nlp_artifacts, [recognizer], [korean_term]
+    )
+
+    assert without[0].score == 0.5
+    assert boosted[0].score == pytest.approx(0.85)
+    assert boosted[0].analysis_explanation.supportive_context_word == korean_term