|
from opencc import OpenCC |
|
cc = OpenCC('t2s') |
|
from h_corpus import Fileset |
|
import unicodedata |
|
import json |
|
def fullwidth_to_halfwidth(input_str): |
|
return ''.join([unicodedata.normalize('NFKC', char) for char in input_str]) |
|
|
|
def clearT(s): |
|
s = cc.convert(fullwidth_to_halfwidth(s)) |
|
return s.lstrip('n').strip().strip(r'\n').replace('\n','\\n') |
|
|
|
a = Fileset(r'D:\datasets\tmp', ext='.jsonl') |
|
b = r'D:\datasets\b-corpus\ChatHaruhi\RolePlaying' |
|
for i in range(len(a)): |
|
path = a[i] |
|
name = path[path.rindex('\\')+1:path.rindex('.')] |
|
print(name) |
|
sc = [] |
|
with open(path, encoding='utf-8') as f: |
|
for line in f: |
|
tmp = json.loads(line) |
|
if tmp['luotuo_openai'] == 'system_prompt': |
|
n = 'system' |
|
d = clearT(tmp['text']) |
|
sc.append(n + '：' + d) |
|
continue |
|
if tmp['luotuo_openai'] == 'config': |
|
continue |
|
tmp = tmp['text'].split('\n') |
|
for one in tmp: |
|
one = one.strip() |
|
if not one: |
|
continue |
|
if ':「' in one: |
|
one = one.replace(':「', '：「', 1) |
|
elif ': 「' in one: |
|
one = one.replace(': 「', '：「', 1) |
|
elif one.startswith(':'): |
|
one = '旁白：' + one[1:] |
|
elif '：「' in one or one.startswith('旁白：'): |
|
pass |
|
elif one.startswith('旁白: '): |
|
one = one.replace('旁白: ', '旁白：', 1) |
|
elif one.startswith('旁白:'): |
|
one = one.replace('旁白:', '旁白：', 1) |
|
elif one.startswith('scene:'): |
|
one = one.replace('scene:', '旁白：', 1) |
|
elif one.startswith('Narrator:'): |
|
one = one.replace('Narrator:', 'Narrator：', 1) |
|
elif one.startswith('春日:'): |
|
one = one.replace('春日:', '春日：', 1) |
|
elif '：' in one: |
|
pass |
|
elif ':' in one: |
|
one = one.replace(':', '：', 1) |
|
else: |
|
one = '旁白：' + one |
|
print(one) |
|
sc.append(one) |
|
if sc: |
|
with open(b + f'\\{name}.txt', 'w', encoding='utf-8') as f: |
|
f.write('\n'.join(sc)) |
|
|
