def load(
cls,
path_to_quickumls: str,
accepted_semtypes: Optional[Set[str]] = None,
threshold: float = 0.9,
similarity_name: str = "jaccard",
spacy_string: str = "en_core_sci_sm",
best_match: bool = False,
n_workers: int = 1,
) -> "QuickUMLSClassifier":
if accepted_semtypes is None:
accepted_semtypes = ALL_SEMTYPES
q = QuickUMLS(path_to_quickumls,
accepted_semtypes=accepted_semtypes,
threshold=threshold,
similarity_name=similarity_name)
# Load the spacy model, Disable the NER and Parser.
q.nlp = spacy.load(spacy_string, disable=("ner", "parser"))
return cls(q, n_workers)
def load(
cls,
path_to_quickumls: str,
accepted_semtypes: Optional[Set[str]] = None,
threshold: float = 0.9,
similarity_name: str = "jaccard",
pooling: str = "mean",
spacy_string: str = "en_core_sci_sm",
priors: Optional[Dict[str, float]] = None,
n_workers: int = 1,
) -> "QuickUMLSClassifier":
"""
Load a QuickUMLSClassifier instance.
:param path_to_quickumls: The path to a valid quickUMLS installation.
:param accepted_semtypes: A set of accepted semantic types. If this is None, we revert to all semantic types.
:param threshold: The threshold to accept.
:param similarity_name: The name of the similarity function. Accepted are 'jaccard', 'overlap', 'cosine' and 'dice'.
:param pooling: The name of the pooling function to use. Should be 'mean', 'max' or 'sum'.
:param spacy_string: The string of the spacy model to use.
:param priors: None or a dictionary mapping from semantic types to class probabilities.
:param n_workers: The number of workers to use during prediction.
:return: An initialized QuickUMLSClassifier.
"""
# Fail early
if pooling not in cls.FUNCS:
raise ValueError(
f"mode should be in {cls.FUNCS}, is now {pooling}")
if accepted_semtypes is None:
accepted_semtypes = ALL_SEMTYPES
q = QuickUMLS(path_to_quickumls,
accepted_semtypes=accepted_semtypes,
threshold=threshold,
similarity_name=similarity_name)
# Load the spacy model, Disable the NER and Parser.
q.nlp = spacy.load(spacy_string, disable=("ner", "parser"))
return cls(q, pooling, priors, n_workers)
