medacy.model.model module¶
A medaCy named entity recognition model wraps together three functionalities
-
class
medacy.model.model.Model(medacy_pipeline=None, model=None, n_jobs=4)[source]¶ Bases:
object-
_extract_features(data_file, medacy_pipeline, is_metamapped)[source]¶ A multi-processed method for extracting features from a given DataFile instance. :param conn: pipe to pass back data to parent process :param data_file: an instance of DataFile :return: Updates queue with features for this given file.
-
cross_validate(num_folds=10)[source]¶ Performs k-fold stratified cross-validation using our model and pipeline. :param num_folds: number of folds to split training data into for cross validation :return: Prints out performance metrics
-
dump(path)[source]¶ Dumps a model into a pickle file :param path: Directory path to dump the model :return:
-
fit(dataset)[source]¶ Runs dataset through the designated pipeline, extracts features, and fits a conditional random field. :param training_data_loader: Instance of Dataset. :return model: a trained instance of a sklearn_crfsuite.CRF model.
-
get_info(return_dict=False)[source]¶ Retrieves information about a Model including details about the feature extraction pipeline, features utilized, and learning model. :param return_dict: Returns a raw dictionary of information as opposed to a formatted string :return: Returns structured information
-
load(path)[source]¶ Loads a pickled model. :param path: File path to directory where fitted model should be dumped :return:
-
static
load_external(package_name)[source]¶ Loads an external medaCy compatible Model. Require’s the models package to be installed Alternatively, you can import the package directly and call it’s .load() method. :param package_name: the package name of the model :return: an instance of Model that is configured and loaded - ready for prediction.
-