def _trainTimex(timexFeatures, timexLabels, grid=False):
"""
Purpose: Train a classifer for Timex3 expressions
@param tokenVectors: A list of tokens represented as feature dictionaries
@param Y: A list of lists of Timex3 classifications for each token in each sentence
"""
assert len(timexFeatures) == len(timexLabels), "{} != {}".format(len(timexFeatures), len(timexLabels))
Y = [l["entity_label"] for l in timexLabels]
clf, vec = train_classifier(timexFeatures, Y, do_grid=grid, ovo=True)
return clf, vec
def _trainEventClass(eventClassFeatures, eventClassLabels, grid=False):
"""
Model::_trainEventClass()
Purpose: Train a classifer for event class identification
@param tokenVectors: A list of tokens represented as feature dictionaries
@param Y: A list of lists of event classifications for each token, with one list per sentence
"""
assert len(eventClassFeatures) == len(eventClassLabels), "{} != {}".format(len(eventClassFeatures), len(eventClassLabels))
Y = [l["entity_label"] for l in eventClassLabels]
clf, vec = train_classifier(eventClassFeatures, Y, do_grid=grid)
return clf, vec
def _trainTlink(tokenVectors, Y, grid=False):
"""
Model::_trainRelation()
Purpose: Train a classifer for temporal relations between events and timex3 labels
@param tokenVectors: A list of tokens represented as feature dictionaries
@param Y: A list of relation classifications for each pair of timexes and events.
"""
print len(tokenVectors)
print len(Y)
assert len(tokenVectors) == len(Y)
clf, vec = train_classifier(tokenVectors, Y, do_grid=grid, ovo=True)
return clf, vec
