fix(analyzer): use the PatternRecognizer default when deny_list_score is omitted in a registry config

[L4XB]

Change Description

A custom deny-list recognizer that omits deny_list_score currently gets a different score depending on how it is loaded: PatternRecognizer(...) and RecognizerRegistry.add_pattern_recognizer_from_dict(...) give the constructor default 1.0, while RecognizerRegistryProvider (YAML / dict registry config) gives 0.0, because CustomRecognizerConfig.deny_list_score declared default=0.0 and that default was dumped and passed to the recognizer. A deny-list match with score 0.0 is filtered by any positive score_threshold, so a YAML deny list without an explicit score never fired through the provider path.

This PR:

• changes CustomRecognizerConfig.deny_list_score to Optional[float] = Field(default=None, ge=0.0, le=1.0);
• gives CustomRecognizerConfig the same model_dump(exclude_none=True) override the other pass-through config models already have, so an omitted field is left out of the kwargs that reach PatternRecognizer.from_dict and the constructor default applies. Other Optional fields on the custom config (patterns, context, supported_language, country_code, ...) already default to None in the recognizer constructor, and the loader reads them with .get(...), so leaving them out of the dump does not change their handling;
• documents the default in docs/analyzer/recognizer_registry_provider.md.

Behavior change (existing users): a registry config (YAML or dict) whose custom recognizer has a deny_list but no deny_list_score now yields deny-list matches with score 1.0 instead of 0.0, i.e. those terms are now detected instead of being dropped below the threshold. Configs that set deny_list_score explicitly are unchanged. Detection patterns and scores of the predefined recognizers are not touched.

Tests:

• model-level: an omitted deny_list_score is None and absent from the dump; an explicit value is dumped unchanged;
• through RecognizerRegistryProvider (new-style entry and legacy supported_language entry): the constructed recognizer has deny_list_score == 1.0, its deny-list pattern has score 1.0, the same as an in-code PatternRecognizer, and analyze("Dear Mr. Smith") returns the match with score 1.0; an explicit deny_list_score: 0.4 is honored.

Ran uv run pytest tests/test_yaml_recognizer_models.py tests/test_recognizer_registry_provider.py (104 passed; the two *_chunker_config tests fail on main too in an environment without the transformers extra) and ruff check / ruff format --check on the changed source file.

Issue reference

Fixes #2242

Checklist

• I have reviewed the contribution guidelines
• I agree to follow this project's Code of Conduct
• I confirm that I have the right to submit this contribution and that it does not knowingly contain proprietary or confidential code.
• My code includes unit tests
• All unit tests and lint checks pass locally
• My PR contains documentation updates / additions if required

[Copilot]

🟢 Approval recommended

The change is narrowly scoped, aligns provider behavior with existing constructor defaults, and is covered by both model-level and provider-path tests.

Pull request overview

This PR fixes a scoring inconsistency for custom deny-list recognizers loaded via RecognizerRegistryProvider by ensuring that an omitted deny_list_score does not get serialized as 0.0 and passed to PatternRecognizer, allowing the constructor default (1.0) to apply consistently across all construction paths.

Changes:

• Update CustomRecognizerConfig.deny_list_score to Optional[float] = None and ensure CustomRecognizerConfig.model_dump() excludes None by default so omitted YAML keys don’t override constructor defaults.
• Add model-level tests to verify omitted vs explicit deny_list_score serialization behavior.
• Add provider-level tests to confirm the registry/provider path yields the same deny-list score behavior as in-code PatternRecognizer construction, and update the provider docs accordingly.

File summaries

File	Description
presidio-analyzer/presidio_analyzer/input_validation/yaml_recognizer_models.py	Makes deny_list_score optional and ensures omitted fields don’t get dumped as explicit kwargs.
presidio-analyzer/tests/test_yaml_recognizer_models.py	Adds unit tests asserting omitted deny_list_score is absent from dumps and explicit values are preserved.
presidio-analyzer/tests/test_recognizer_registry_provider.py	Adds integration tests verifying provider-loaded custom deny-list recognizers use the PatternRecognizer default score when omitted, and honor explicit scores.
docs/analyzer/recognizer_registry_provider.md	Documents that deny_list_score is optional and defaults to PatternRecognizer’s 1.0 when omitted.

Review details

• Files reviewed: 4/4 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.