feedback_forensics.tools.ff_human_annotation#
Module Contents#
Functions#
Ensure that an annotator entry exists for every trait. |
|
Extract displayable text for a single response from AnnotatedPairs. |
|
Return (prompt, text_a, text_b) for a comparison in v1.0 or v2.0. |
|
Map radio value to AnnotatedPairs ‘pref’ string. |
|
Inverse mapping to populate radio controls from existing annotations. |
|
Data#
API#
- feedback_forensics.tools.ff_human_annotation.PERSONALITY_TRAITS_DEFAULT: List[str]#
[‘is more verbose’, ‘has more structured formatting’, ‘makes more confident statements’, ‘is more fa…
- feedback_forensics.tools.ff_human_annotation._ensure_trait_annotators_exist(ap: Dict[str, Any], traits: List[str]) Dict[str, str]#
Ensure that an annotator entry exists for every trait.
Returns mapping from trait -> annotator_id (stable, hashed).
- feedback_forensics.tools.ff_human_annotation._get_text_from_response(response: Any) str#
Extract displayable text for a single response from AnnotatedPairs.
Supports both v1.0 (string fields) and v2.0 (dict response fields).
- feedback_forensics.tools.ff_human_annotation._read_pair_texts(comparison: Dict[str, Any]) Tuple[str, str, str]#
Return (prompt, text_a, text_b) for a comparison in v1.0 or v2.0.
- feedback_forensics.tools.ff_human_annotation._annotation_from_value(value: str) str#
Map radio value to AnnotatedPairs ‘pref’ string.
Radio values: ‘text_a’, ‘text_b’, ‘not relevant’
- feedback_forensics.tools.ff_human_annotation._value_from_annotation(pref: str) str#
Inverse mapping to populate radio controls from existing annotations.
- feedback_forensics.tools.ff_human_annotation._save(ap: Dict[str, Any], output_path: pathlib.Path) None#
- feedback_forensics.tools.ff_human_annotation.build_interface(input_path: pathlib.Path, output_path: pathlib.Path | None, traits: List[str] | None = None, use_standard_traits: bool = False) gradio.Blocks#
- feedback_forensics.tools.ff_human_annotation.run()#