def test_check_consistent_bilou_tagging(
tags: List[Text],
expected_tags: List[Text],
debug_message: Optional[Text],
caplog: LogCaptureFixture,
):
with caplog.at_level(logging.DEBUG):
actual_tags = bilou_utils.ensure_consistent_bilou_tagging(tags)
if debug_message:
assert len(caplog.records) > 0
assert debug_message in caplog.text
else:
assert len(caplog.records) == 0
assert actual_tags == expected_tags
def _tag_confidences(
self, tokens: List[Token], predictions: Dict[Text, List[Dict[Text, float]]]
) -> Tuple[Dict[Text, List[Text]], Dict[Text, List[float]]]:
"""Get most likely tag predictions with confidence values for tokens."""
tags = {}
confidences = {}
for tag_name, predicted_tags in predictions.items():
if len(tokens) != len(predicted_tags):
raise Exception(
"Inconsistency in amount of tokens between crfsuite and message"
)
_tags, _confidences = self._most_likely_tag(predicted_tags)
if self.component_config[BILOU_FLAG]:
_tags = bilou_utils.ensure_consistent_bilou_tagging(_tags)
confidences[tag_name] = _confidences
tags[tag_name] = _tags
return tags, confidences
