bot-yaya commited on
Commit
50f74f4
1 Parent(s): 13e4794

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +45 -1
README.md CHANGED
@@ -23,4 +23,48 @@ configs:
23
  ---
24
  # 联合国语料中文的机翻英文
25
 
26
- rt,翻译工具用的argostranslate,因为对齐之前需要做一次机翻,所以这里上传了一份,以便后续pipeline使用,其它语言也一样。注意不要直接拿这份去练机翻模型,因为它们不是人翻的。
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
  ---
24
  # 联合国语料中文的机翻英文
25
 
26
+ rt,翻译工具用的argostranslate,因为对齐之前需要做一次机翻,所以这里上传了一份,以便后续pipeline使用,其它语言也一样。注意不要直接拿这份去练机翻模型,因为它们不是人翻的。
27
+
28
+ 在机翻之前,已经用脚本洗掉了一部分制表噪声和分隔符,所用函数如下:
29
+
30
+ ```
31
+ def clean_paragraph(paragraph):
32
+ lines = paragraph.split('\n')
33
+ para = ''
34
+ table = []
35
+
36
+ for line in lines:
37
+ line = line.strip()
38
+
39
+ # 表格线或其他分割线
40
+ if re.match(r'^\+[-=+]+\+|-+|=+|_+$', line):
41
+ if not para.endswith('\n'):
42
+ para += '\n'
43
+ if len(table) > 0:
44
+ para += '\t'.join(table)
45
+ table = []
46
+ # 表格中的空行
47
+ elif re.match(r'^\|( +\|)+$', line):
48
+ para += '\t'.join(table) + ' '
49
+ table = []
50
+ # 表格中的内容行
51
+ elif re.match(r'^\|([^|]+\|)+$', line):
52
+ if len(table) == 0:
53
+ table = line[1:-2].split('|')
54
+ else:
55
+ arr = line[1:-2].split('|')
56
+ if len(arr) == len(table):
57
+ table = [table[i].strip() + arr[i].strip() for i in range(len(table))]
58
+ elif len(arr) > len(table):
59
+ table = [table[i].strip() + arr[i].strip() if i < len(table) else arr[i].strip() for i in range(len(arr))]
60
+ else:
61
+ table = [table[i].strip() + arr[i].strip() if i < len(arr) else table[i].strip() for i in range(len(table))]
62
+ # 正文内容
63
+ else:
64
+ para += ' ' + line
65
+ if len(table) > 0:
66
+ if not para.endswith('\n'):
67
+ para += '\n'
68
+ para += '\t'.join(table)
69
+ return re.sub(r'[ \t]{2,}', ' ', re.sub(r'\n{2,}', '\n', para)).strip()
70
+ ```