def load_data_for_segmentation(doc_num):
print('Interview:', doc_num)
path = './data/segmentation/sentence/interview-text_' + doc_num + '.txt'
# ans
# path = './data/eval/interview-text_sentence_' + doc_num + '.txt'
return utils.load_data_segment(path)
def load_data_for_segmentation(doc_num, *, ans=False):
print('Interview:', doc_num)
path = './data/segmentation/sentence/interview-text_' + doc_num + '.txt'
# path = './data/segmentation/utterance/interview-text_' + doc_num + '.txt'
if ans:
path = './data/eval/interview-text_sentence_' + doc_num + '.txt'
return utils.load_data_segment(path)
def evaluation(res, segmentation_model, segmentation_type, model_type,
doc_type, doc_num):
count = 0
label_for_eval = []
path_for_eval = './data/eval/interview-text_' + doc_type + '_' + doc_num + '.txt'
sentence = False
if doc_type == 'sentence':
sentence = True
data_for_eval = utils.load_data_segment(path_for_eval, sentence=sentence)
label_for_eval = [
item[0] for item in data_for_eval.items() if '____' in item[1][0]
]
print('予測:', res.index.values)
print('正解:', label_for_eval)
# 精度、再現率
plus_label_for_eval = [val + 1 for val in label_for_eval]
minas_label_for_eval = [val - 1 for val in label_for_eval]
approx_label_for_eval = copy.deepcopy(label_for_eval)
approx_label_for_eval.extend(plus_label_for_eval)
approx_label_for_eval.extend(minas_label_for_eval)
for i in res.index.values:
i = float(i)
# 前後許容
if i in approx_label_for_eval or i + 1 in approx_label_for_eval or i - 1 in approx_label_for_eval:
count += 1
p = count / len(res.index.values)
r = count / len(label_for_eval)
f_score = 2 * 1 / (1 / r + 1 / p)
print("Precision:", p)
print("Recall:", r)
print("F score:", f_score)
save_path = './result/segmentation/' + segmentation_type + '/' + model_type + '/' + doc_type + '/evaluation/' + 'doc_num_' + doc_num + '_' + model_type + '_window_size_' + str(
segmentation_model.window_size) + '_' + str(datetime.date.today())
with open(save_path + '.txt', 'w') as f:
print('予測:', res.index.values, file=f)
print('正解:', label_for_eval, file=f)
print("Precision:", p, file=f)
print("Recall:", r, file=f)
print("F score:", f_score, file=f)
return count, label_for_eval, f_score