-
Notifications
You must be signed in to change notification settings - Fork 1
/
dataset.py
41 lines (38 loc) · 1.49 KB
/
dataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
import json
import csv
import re
import glob
class ConversationDataset():
'''
The conversation database class.
'''
def __init__(self, path_to_dataset, batch_size, max_size):
self.batches = []
self.max_len = 512
print("Reading data from", path_to_dataset, "batch size", batch_size)
all_data_list = glob.glob(path_to_dataset + '*')
all_data_list.sort()
all_data_list = all_data_list[:max_size] # max size
files_in_batch = 0
for data_file in all_data_list:
f = open(data_file)
data = f.readlines()
data = [d.strip() for d in data]
data_id = data_file.split('/')[-1]
if files_in_batch == 0:
self.batches.append({'conversations':{}, 'responses_pool':[], 'answers_pool':[]})
self.batches[-1]['conversations'][data_id] = data
for ut_num in range(len(data)):
if ut_num % 2 and ut_num != (len(data) - 1) :
self.batches[-1]['responses_pool'].append(data[ut_num])
self.batches[-1]['answers_pool'].append(data[-1])
files_in_batch += 1
if files_in_batch == batch_size:
files_in_batch = 0
'''
def processing(input_text, max_len):
result = re.sub('\s\s+', ' ', input_text) # replace more than one whitespace with one.
result = re.sub(r'http\S+', '', result, flags=re.MULTILINE) # remove URL
result = result[:max_len]
return result.strip()
'''