def get_match_result(true_seqs, pred_seqs, do_stem=True, type='exact'):
'''
:param true_seqs:
:param pred_seqs:
:param do_stem:
:param topn:
:param type: 'exact' or 'partial'
:return:
'''
micro_metrics = []
micro_matches = []
# do processing to baseline predictions
match_score = np.asarray([0.0] * len(pred_seqs), dtype='float32')
target_number = len(true_seqs)
predicted_number = len(pred_seqs)
metric_dict = {'target_number': target_number, 'prediction_number': predicted_number, 'correct_number': match_score}
# convert target index into string
if do_stem:
true_seqs = [stem_word_list(seq) for seq in true_seqs]
pred_seqs = [stem_word_list(seq) for seq in pred_seqs]
for pred_id, pred_seq in enumerate(pred_seqs):
if type == 'exact':
match_score[pred_id] = 0
for true_id, true_seq in enumerate(true_seqs):
match = True
if len(pred_seq) != len(true_seq):
continue
for pred_w, true_w in zip(pred_seq, true_seq):
# if one two words are not same, match fails
if pred_w != true_w:
match = False
break
# if every word in pred_seq matches one true_seq exactly, match succeeds
if match:
match_score[pred_id] = 1
break
elif type == 'partial':
max_similarity = 0.
pred_seq_set = set(pred_seq)
# use the jaccard coefficient as the degree of partial match
for true_id, true_seq in enumerate(true_seqs):
true_seq_set = set(true_seq)
jaccard = len(set.intersection(*[set(true_seq_set), set(pred_seq_set)])) / float(len(set.union(*[set(true_seq_set), set(pred_seq_set)])))
if jaccard > max_similarity:
max_similarity = jaccard
match_score[pred_id] = max_similarity
elif type == 'bleu':
# account for the match of subsequences, like n-gram-based (BLEU) or LCS-based
match_score[pred_id] = bleu(pred_seq, true_seqs, [0.1, 0.3, 0.6])
return match_score
def get_match_result(true_seqs, pred_seqs, do_stem=True, type='exact'):
'''
:param true_seqs:
:param pred_seqs:
:param do_stem:
:param topn:
:param type: 'exact' or 'partial'
:return:
'''
micro_metrics = []
micro_matches = []
# do processing to baseline predictions
match_score = np.asarray([0.0] * len(pred_seqs), dtype='float32')
target_number = len(true_seqs)
predicted_number = len(pred_seqs)
metric_dict = {'target_number': target_number, 'prediction_number': predicted_number, 'correct_number': match_score}
# convert target index into string
if do_stem:
true_seqs = [stem_word_list(seq) for seq in true_seqs]
pred_seqs = [stem_word_list(seq) for seq in pred_seqs]
for pred_id, pred_seq in enumerate(pred_seqs):
if type == 'exact':
match_score[pred_id] = 0
for true_id, true_seq in enumerate(true_seqs):
match = True
if len(pred_seq) != len(true_seq):
continue
for pred_w, true_w in zip(pred_seq, true_seq):
# if one two words are not same, match fails
if pred_w != true_w:
match = False
break
# if every word in pred_seq matches one true_seq exactly, match succeeds
if match:
match_score[pred_id] = 1
break
elif type == 'partial':
max_similarity = 0.
pred_seq_set = set(pred_seq)
# use the jaccard coefficient as the degree of partial match
for true_id, true_seq in enumerate(true_seqs):
true_seq_set = set(true_seq)
jaccard = len(set.intersection(*[set(true_seq_set), set(pred_seq_set)])) / float(len(set.union(*[set(true_seq_set), set(pred_seq_set)])))
if jaccard > max_similarity:
max_similarity = jaccard
match_score[pred_id] = max_similarity
elif type == 'bleu':
# account for the match of subsequences, like n-gram-based (BLEU) or LCS-based
match_score[pred_id] = bleu(pred_seq, true_seqs, [0.1, 0.3, 0.6])
return match_score