发表于 2019-01-02   |   标签 Python   |  

翻译遇到麻烦的地方，就是人物名词、地名、拟声词还有专有名词，这些在中文里面有些是音译（比如地名、人名），如果直接用翻译软件往往会翻出一些奇怪的东西来。这时候就需要用到字典，其实就变成来说就是映射，将符合条件的原文字符串不翻译，直接映射出原文字符串。

打开Jbeijng-》加载字典-》编辑字典

然后找到对应的字典文件

接下来到vnr的代码里面看下是如何加载

  def openAllUserDic(self, paths):
    """Add one user dictionary.
    @param  paths  [unicode]  path to Jcuser.dic, but without ".dic" suffix.
                              At most 3 elements.
    @return  bool
    @raise  WindowsError, AttributeError

    Guessed from OllyDbg:
    //int __cdecl DJC_OpenAllUserDic_Unicode(LPWSTR, LPWSTR unknown, LPWSTR unknown, int unknown)
    int __cdecl DJC_OpenAllUserDic_Unicode(LPWSTR, int unknown)
    Return 1 or -255 if succeeded.

    According to how it is invoked in JCT.exe (push 0), unknown is always 0.
    This function will beep when failed.

    According to how DJC_OpenAllUserDic_Unicode invoke DJC_OpenAllUserDic,
    the first parameter type is supposed to be:
        wchar_t[0x408/sizeof(wchar_t)][3]

    10020150 >/$ 81EC 10060000  SUB ESP,0x610
    10020156  |. A1 50E70710    MOV EAX,DWORD PTR DS:[0x1007E750]
    1002015B  |. 33C4           XOR EAX,ESP
    1002015D  |. 898424 0C06000>MOV DWORD PTR SS:[ESP+0x60C],EAX
    10020164  |. 55             PUSH EBP
    10020165  |. 8B2D A4F20610  MOV EBP,DWORD PTR DS:[<&KERNEL32.WideCha>;  kernel32.WideCharToMultiByte
    1002016B  |. 56             PUSH ESI
    1002016C  |. 8BB424 1C06000>MOV ESI,DWORD PTR SS:[ESP+0x61C]
    10020173  |. 6A 00          PUSH 0x0                                 ; /pDefaultCharUsed = NULL
    10020175  |. 6A 00          PUSH 0x0                                 ; |pDefaultChar = NULL
    10020177  |. 68 04020000    PUSH 0x204                               ; |MultiByteCount = 204 (516.) ; jichi 12/31/2013: 0x204 comes from here
    1002017C  |. 8D4424 14      LEA EAX,DWORD PTR SS:[ESP+0x14]          ; |
    10020180  |. 50             PUSH EAX                                 ; |MultiByteStr
    10020181  |. 6A FF          PUSH -0x1                                ; |WideCharCount = FFFFFFFF (-1.)
    10020183  |. 56             PUSH ESI                                 ; |WideCharStr
    10020184  |. 6A 00          PUSH 0x0                                 ; |Options = 0
    10020186  |. 6A 00          PUSH 0x0                                 ; |CodePage = CP_ACP
    10020188  |. FFD5           CALL EBP                                 ; \WideCharToMultiByte
    1002018A  |. 8BC6           MOV EAX,ESI
    1002018C  |. 8D50 02        LEA EDX,DWORD PTR DS:[EAX+0x2]
    1002018F  |. 90             NOP

    """
    if not paths:
      return False
    if len(paths) > MAX_USERDIC_COUNT:
      print("too many user-defined dictionaries")

    MAX_PATH_LENGTH = USERDIC_PATH_SIZE

    #path = os.path.splitext(path)[0] # remove ".dic" suffix
    buf = self.userdicBuffer
    ctypes.memset(buf, 0, USERDIC_BUFFER_SIZE) # zero memory

    for i in xrange(min(len(paths), MAX_USERDIC_COUNT)):
      path = paths[i]
      if len(path) > MAX_PATH_LENGTH:
        print("path is too long: %s" % path)
        continue
      offset = i * USERDIC_PATH_SIZE
      for index, c in enumerate(path):
        buf[index + offset] = c

    ret = self.dll.DJC_OpenAllUserDic_Unicode(buf, 0)
    return ret in (1,-255)

  #未使用字典
  ret = Loader().translate(u"姫",simplified=True)
  print ret
  ret = Loader().translate(u"ＷＴＦ",simplified=True)
  print ret

  #加载字典
  l = Loader()
  l.setUserDic((
    u"Y:\JDict\China",
  ))
  ret = Loader().translate(u"姫",simplified=True)
  print ret
  ret = Loader().translate(u"ＷＴＦ",simplified=True)
  print ret

J北京对流行语，网络用语之类的翻译能力较差，这种时候就需要一个好的字典了，网上都有类似的字典，大部分都用json来存储，下阶段研究下如果通过python自动添加字典。