def test(embeddingsFile, existingModel, predictionsFile,
documents_are_sequences, sentenceCNN, charCNN, uniLSTM, useBERT):
embeddings, vocab = event_reader.load_embeddings(embeddingsFile,
vocab_size,
word_embedding_dim)
testSentences, testBookIndex = event_reader.prepare_annotations_from_folder(
test_folder, documents_are_sequences, useBERT)
testC, testX, testP, testW, testY, testL = transform_examples(
testSentences, vocab, useBERT)
test_generator = single_generator(testC, testX, testP, testW, testY, testL)
test_metadata = event_reader.convert_to_index(testSentences)
model = event_cnn(embeddings, sentenceCNN, charCNN, uniLSTM, useBERT)
model.load_weights(existingModel)
predictionFile = predictionsFile
out = open(predictionFile, "w", encoding="utf-8")
gold = []
preds = []
c = 0
for step in range(len(testL)):
batch, y = next(test_generator)
probs = model.predict_on_batch(batch)
_, length, _ = y.shape
for i in range(length):
out.write("%s\t%s\t%s\t%.20f\n" %
('\t'.join([str(x) for x in test_metadata[c]]),
int(probs[0][i][0] > 0.5), y[0][i][0], probs[0][i][0]))
preds.append(probs[0][i][0] >= 0.5)
gold.append(y[0][i][0])
c += 1
f, p, r, correct, trials, trues = event_eval.check_f1_two_lists(
gold, preds)
print("precision: %.3f %s/%s" % (p, correct, trials))
print("recall: %.3f %s/%s" % (r, correct, trues))
print("F: %.3f" % f)
event_eval.check_f1_two_lists(gold, preds)
out.close()
def predict_file_batch(filename, embeddingsFile, existingModel,
predictionsFile, documents_are_sequences, sentenceCNN,
charCNN, pad, uniLSTM, useBERT):
embeddings, vocab = event_reader.load_embeddings(embeddingsFile,
vocab_size,
word_embedding_dim)
testSentences, testBookIndex = event_reader.prepare_annotations_from_file(
filename,
documents_are_sequences=documents_are_sequences,
useBERT=useBERT)
testC, testX, testP, testW, testY, testL = transform_examples(
testSentences, vocab, useBERT)
test_metadata = event_reader.convert_to_index(testSentences)
max_sequence_length = 0
for length in testL:
if length > max_sequence_length:
max_sequence_length = length
print("max l: ", max_sequence_length)
model = event_cnn(embeddings, sentenceCNN, charCNN, uniLSTM, useBERT)
model.load_weights(existingModel)
predictionFile = predictionsFile
out = open(predictionFile, "w", encoding="utf-8")
if pad:
testC, testX, testP, testW, testY, testL = pad_all(
testC, testX, testP, testW, testY, testL, max_sequence_length,
useBERT)
model.predict([testC, testW, testX, testP])
probs = model.predict([testC, testW, testX, testP], batch_size=128)
c = 0
lastSent = None
for step in range(len(testL)):
for i in range(testL[step]):
sid = test_metadata[c][1]
if lastSent != sid and lastSent != None:
out.write("\n")
w_book, w_sid, w_word, _ = test_metadata[c]
label = "O"
if probs[step][i][0] > 0.5:
label = "EVENT"
out.write("%s\t%s\n" % (w_word, label))
lastSent = sid
c += 1
else:
test_generator = single_generator(testC, testX, testP, testW, testY,
testL)
c = 0
for step in range(len(testL)):
batch, y = next(test_generator)
probs = model.predict_on_batch(batch)
_, length, _ = y.shape
for i in range(length):
sid = test_metadata[c][1]
w_book, w_sid, w_word, _ = test_metadata[c]
label = "O"
if probs[0][i][0] > 0.5:
label = "EVENT"
out.write("%s\t%s\n" % (w_word, label))
c += 1
out.write("\n")
out.close()
if __name__ == "__main__":
outputFile = sys.argv[1]
nlp = spacy.load('en', disable=['ner,parser'])
nlp.remove_pipe('ner')
nlp.remove_pipe('parser')
train_folder = "../data/bert/train"
dev_folder = "../data/bert/dev"
test_folder = "../data/bert/test"
testSentences, _ = event_reader.prepare_annotations_from_folder(
test_folder)
test_metadata = event_reader.convert_to_index(testSentences)
golds = []
preds = []
for sentence in testSentences:
tokens_list = [word[0] for word in sentence]
tokens = nlp.tokenizer.tokens_from_list(tokens_list)
nlp.tagger(tokens)
for idx, token in enumerate(tokens):
pred = 0
if token.tag_.startswith("V"):
pred = 1
preds.append(pred)
label = sentence[idx][1]
golds.append(label)