def load_event_classifier(event_type):
classifier_file = MODEL_DIR + event_type + EVENT_DETECT_MODEL_SUFFIX
feature_map_file = MODEL_DIR + event_type + EVENT_DETECT_FEATMAP_SUFFIX
classifier, feature_map = Classification.load_classifier(
classifier_file, feature_map_file)
return classifier, feature_map
def load_event_filling_classifier():
classifier_file = MODEL_DIR + EVENT_FILLER_MODEL_NAME
feature_map_file = MODEL_DIR + EVENT_FILLER_FEATMAP_NAME
classifier, feature_map = Classification.load_classifier(
classifier_file, feature_map_file)
return classifier, feature_map
def load_classifier(event_type):
classifier_file = MODEL_DIR + event_type + STATUS_CLASSF_MODEL_SUFFIX
feature_map_file = MODEL_DIR + event_type + STATUS_CLASSF_FEATMAP_SUFFIX
classifier, feature_map = Classification.load_classifier(classifier_file, feature_map_file)
return classifier, feature_map
def load_classifier(event_type):
classifier_file = MODEL_DIR + event_type + STATUS_CLASSF_MODEL_SUFFIX
feature_map_file = MODEL_DIR + event_type + STATUS_CLASSF_FEATMAP_SUFFIX
classifier, feature_map = Classification.load_classifier(
classifier_file, feature_map_file)
return classifier, feature_map
def classify_sent_for_substance(classifier, feature_map, sent, substance, sentences_predicted_to_have_events):
sent_feats = get_features(sent)
classifications = Classification.classify_instance(classifier, feature_map, sent_feats)
# Add detected event to sentence
if classifications[0] == HAS_SUBSTANCE:
event = Event(substance)
sent.predicted_events.append(event)
sentences_predicted_to_have_events.append(sent)
def train_event_fillers(patients):
# Train model
features, labels = Processing.features_and_labels(patients)
classifier, feature_map = Classification.train_classifier(features, labels)
# Write models to file
classf_file = MODEL_DIR + EVENT_FILLER_MODEL_NAME
featmap_file = MODEL_DIR + EVENT_FILLER_FEATMAP_NAME
joblib.dump(classifier, classf_file)
Pickle.dump(feature_map, open(featmap_file, "wb"))
def classify_sent_for_substance(classifier, feature_map, sent, substance,
sentences_predicted_to_have_events):
sent_feats = get_features(sent)
classifications = Classification.classify_instance(classifier, feature_map,
sent_feats)
# Add detected event to sentence
if classifications[0] == HAS_SUBSTANCE:
event = Event(substance)
sent.predicted_events.append(event)
sentences_predicted_to_have_events.append(sent)
def add_attrib_using_classifier(classifier, feature_map, features, attribs_found_per_substance, attrib):
""" For an attribute that can be different substances, using ML classifier """
classifications = Classification.classify_instance(classifier, feature_map, features)
# Assign attribute to substance identified
classified_substance = classifications[0]
if classified_substance in SUBSTANCE_TYPES:
if attrib.type not in attribs_found_per_substance[classified_substance]:
attribs_found_per_substance[classified_substance][attrib.type] = []
attribs_found_per_substance[classified_substance][attrib.type].append(attrib)
def add_attrib_using_classifier(classifier, feature_map, features,
attribs_found_per_substance, attrib):
""" For an attribute that can be different substances, using ML classifier """
classifications = Classification.classify_instance(classifier, feature_map,
features)
# Assign attribute to substance identified
classified_substance = classifications[0]
if classified_substance in SUBSTANCE_TYPES:
if attrib.type not in attribs_found_per_substance[
classified_substance]:
attribs_found_per_substance[classified_substance][attrib.type] = []
attribs_found_per_substance[classified_substance][attrib.type].append(
attrib)
def train_event_detectors(patients):
# Retrieve features and labels per every sentence
sent_feat_dicts, labels_per_subst = sentence_features_and_labels(patients)
# Get Florian's data to reinforce training
flor_sent_feat_dicts, flor_labels_per_subst = flor_sentence_features_and_labels()
sent_feat_dicts.extend(flor_sent_feat_dicts)
labels_per_subst["Tobacco"].extend(flor_labels_per_subst["Tobacco"])
labels_per_subst["Alcohol"].extend(flor_labels_per_subst["Alcohol"])
#labels_per_subst["Secondhand"].extend(flor_labels_per_subst["Secondhand"])
for substance_type in ML_CLASSIFIER_SUBSTANCES:
# Train classifier
classifier, feature_map = Classification.train_classifier(sent_feat_dicts, labels_per_subst[substance_type])
# Save to file
classf_file = MODEL_DIR + substance_type + EVENT_DETECT_MODEL_SUFFIX
featmap_file = MODEL_DIR + substance_type + EVENT_DETECT_FEATMAP_SUFFIX
joblib.dump(classifier, classf_file)
Pickle.dump(feature_map, open(featmap_file, "wb"))
def load_event_filling_classifier():
classifier_file = MODEL_DIR + EVENT_FILLER_MODEL_NAME
feature_map_file = MODEL_DIR + EVENT_FILLER_FEATMAP_NAME
classifier, feature_map = Classification.load_classifier(classifier_file, feature_map_file)
return classifier, feature_map
def load_event_classifier(event_type):
classifier_file = MODEL_DIR + event_type + EVENT_DETECT_MODEL_SUFFIX
feature_map_file = MODEL_DIR + event_type + EVENT_DETECT_FEATMAP_SUFFIX
classifier, feature_map = Classification.load_classifier(classifier_file, feature_map_file)
return classifier, feature_map