for tgt, pred, topic in zip(tgt_fr, pred_fr, topic_fr):
topic = eval(topic)
tgt = tgt.strip("\n")
tgt = tgt.replace("<SOS> ", "")
tgt = tgt.replace(" <EOS>", "")
pred = pred.strip("\n").split("\t")[0]
# print(pred)
pred = pred.replace("<SOS> ", "")
pred = pred.replace(" <EOS>", "")
for k, v in topic.items():
tgt = tgt.replace(k, v)
pred = pred.replace(k, v)
pred_tokens = pred.strip().split(" ")
gold_tokens = tgt.strip().split(" ")
sents.append([pred_tokens, gold_tokens])
# calc f1
f1 = calc_f1(sents)
# calc bleu
bleu1, bleu2 = calc_bleu(sents)
# calc distinct
distinct1, distinct2 = calc_distinct(sents)
score = f1 + bleu1 + bleu2
output_str = "Score: %.2f%%\n" % (score * 100)
output_str += "F1: %.2f%%\n" % (f1 * 100)
output_str += "BLEU1: %.3f%%\n" % bleu1
output_str += "BLEU2: %.3f%%\n" % bleu2
output_str += "DISTINCT1: %.3f%%\n" % distinct1
output_str += "DISTINCT2: %.3f%%\n" % distinct2
logger.info(output_str)
def compute_distinct2(p, s):
d1, d2 = calc_distinct([[p.split(), s.split()]])
return d2