def shuffle_examples(labels, sentences, pct_train=0.5):
"""
Shuffles training examples and splits into trainin and validation sets.
labels, sentences - File names pointing to data sets.
pct_train - Pctge of data to keep at training data. The rest
becomes validation data.
"""
# first import the training data
train_labels_ordered = dp.labels_as_ints(labels)
train_sentences_ordered = dp.import_sentences(sentences)
Nex = len(train_labels_ordered)
Ntrain = int(pct_train*Nex + 1)
# split into training and validation sets
train_labels = []
train_sentences = []
validation_labels = []
validation_sentences = []
np.random.seed(1987)
examples = np.arange(0,Nex,dtype='int')
np.random.shuffle(examples)
train_examples = examples[:Ntrain]
validation_examples = examples[Ntrain:]
for tex in train_examples:
train_labels.append(train_labels_ordered[tex])
train_sentences.append(train_sentences_ordered[tex])
for vex in validation_examples:
validation_labels.append(train_labels_ordered[vex])
validation_sentences.append(train_sentences_ordered[vex])
return train_labels, train_sentences, validation_labels, validation_sentences
dummy = 0
# load weights
if method != 'dummy':
weights_0p25 = np.load('training_output/'+method+'_w_0p25.npy')
weights_0p5 = np.load('training_output/'+method+'_w_0p5.npy')
weights_0p75 = np.load('training_output/'+method+'_w_0p75.npy')
else:
weights_0p25 = np.load('training_output/collins_w_0p25.npy')
weights_0p5 = np.load('training_output/collins_w_0p5.npy')
weights_0p75 = np.load('training_output/collins_w_0p75.npy')
weights = [weights_0p25, weights_0p5, weights_0p75]
# load the test data
test_labels = dp.labels_as_ints('dataset/testLabels.txt')
test_sentences = dp.import_sentences('dataset/testSentences.txt')
# calculate the scores
score = []
for w in weights:
if scoretype == 'word':
score.append(sr.score_by_word(w,test_labels,test_sentences,dummy))
elif scoretype == 'sentence':
score.append(sr.score_by_sentence(w,test_labels,test_sentences,dummy))
elif scoretype == 'mark':
score.append(sr.score_by_mark(w,test_labels,test_sentences,dummy))
else:
print 'Not a valid method!\n'
exit(0)
