CWrapEngine::CWrapEngine()
{
CheckLanguage();
}
bool CGrammarItem::AddAttribute(string Name, string Value, MorphLanguageEnum Language, string& ErrorStr)
{
if (Value.length() > 0)
if (Value[0] == '"')
{
if ( (Value.length()<2) || (Value[Value.length() - 1] != '"'))
{
ErrorStr = Format("no matching quotation mark for attribute value \"%s\"",Value.c_str());
return false;
};
Value = Value.substr(1, Value.length()-2);
};
if (Name == "root")
{
m_bSynMain = true;
return true;
};
if (Name == "type")
{
m_TokenType = StringToTokenType(Value);
if (m_TokenType == OTHER_TOKEN_TYPE)
{
ErrorStr = Format("unknown token type:%s ",Value.c_str());
return false;
}
};
if (Name == "hom")
{
if (Value == "yes")
m_bCanHaveManyHomonyms = true;
else
if (Value == "no")
m_bCanHaveManyHomonyms = false;
else
{
ErrorStr = Format("Bad value for attribute \"hom\" (\"%s\"). It can be \"yes\" or \"no\"",Value.c_str());
return false;
};
if (m_TokenType == OTHER_TOKEN_TYPE)
m_TokenType = (Language == morphRussian) ? RLE : LLE;
return true;
};
if (Name == "grm")
{
m_MorphPattern.m_GrmAttribute = Value;
if (m_TokenType == OTHER_TOKEN_TYPE)
m_TokenType = (Language == morphRussian) ? RLE : LLE;
return true;
};
if (Name == "form")
{
m_Token = Value;
RmlMakeUpper(m_Token, Language);
m_ItemStrId = Value;
if ( (m_TokenType == OTHER_TOKEN_TYPE) && !m_Token.empty())
{
if (ispunct((BYTE)m_Token[0]))
m_TokenType = PUNCTUAT;
else
if (isdigit((BYTE)m_Token[0]))
m_TokenType = NUM;
else
if (Language == morphRussian)
{
if (CheckLanguage(m_Token, Language))
m_TokenType = RLE;
}
else
{
if (CheckLanguage(m_Token, Language))
m_TokenType = LLE;
}
};
return true;
};
if (Name == "register")
{
if (Value == "AA")
m_Register = UpUp;
else
if (Value == "aa")
m_Register = LowLow;
else
if (Value == "Aa")
m_Register = UpLow;
else
{
ErrorStr = Format("Bad value for attribute \"register\" (\"%s\"). It can be \"AA\", \"aa\" or \"Aa\"",Value.c_str());
return false;
};
if (m_TokenType == OTHER_TOKEN_TYPE)
m_TokenType = (Language == morphRussian) ? RLE : LLE;
return true;
};
if (Name == "filename")
{
Value = GetPathByFile(CurrentSourceFileName) + Value;
if (m_TokenType == OTHER_TOKEN_TYPE)
m_TokenType = (Language == morphRussian) ? RLE : LLE;
}
m_Attributes[Name] = Value;
return true;
};
CDictionarySearch CTrigramModel::find_word(const string& WordStr) const
{
CDictionarySearch R;
assert (!WordStr.empty());
if (WordStr.empty())
{
//fprintf (stderr, "Empty word!\n");
R.m_pFoundWord = 0;
for (WORD i=0; i < m_TagsCount; i++)
R.m_PossibleWordTags.insert(i);
return R;
}
R.m_pFoundWord = lookup_word(WordStr);
if (! R.m_pFoundWord )
{
// если слова нет в словаре, тогда попробуем его поискать в нижнем регистре
string lower = WordStr;
RmlMakeLower(lower, m_Language);
R.m_pFoundWord = lookup_word(lower);
}
if ( R.m_pFoundWord )
{
// приписываем все тэги, которые были в корпусе
for (size_t i=0; i < R.m_pFoundWord->m_Length; i++)
{
int Tag = m_LexProbs[R.m_pFoundWord->m_StartOffset + i].m_Tag;
R.m_PossibleWordTags.insert(Tag);
}
}
// получаем все возможные тэги из морф. словар¤
map<string, const vector<CXmlMorphAnnot>* >::iterator it = m_CurrentSentenceWords2Annots.find(WordStr);
if (it != m_CurrentSentenceWords2Annots.end())
get_tags_from_annots(*it->second,R.m_PossibleWordTags, WordStr);
#ifdef USE_TRIGRAM_LEMMATIZER
else
get_tags_from_lemmatizer_but_not_preps(WordStr,R.m_PossibleWordTags);
#endif
if (R.m_PossibleWordTags.empty())
if ( atoi(WordStr.c_str()) > 0
&& (m_Language==morphRussian)
)
{
for (size_t i=0; i < m_RegisteredTags.size();i++)
if (m_RegisteredTags[i].length() > 3 && m_RegisteredTags[i].substr(0,4) == "„»—Ћ")
R.m_PossibleWordTags.insert(i);
if (R.m_PossibleWordTags.empty())
throw CExpc ("Cannot find „»—Ћ tag");
}
else
if ( ispunct((BYTE)WordStr[0])
|| !CheckLanguage(WordStr,m_Language)
)
{
int tag = find_tag("UNK");
if (tag == UnknownTag)
throw CExpc ("Cannot find UNK tag");
R.m_PossibleWordTags.insert(tag);
}
else
{
// приписываем все тэги
if (!m_bQuiet)
fprintf (stderr, "No information for word %s\n",WordStr.c_str());
for (size_t i=0; i < min((size_t)200, m_TagsOrderedByUnigrams.size()); i++)
{
WORD tagno = m_TagsOrderedByUnigrams[i];
string tag = m_RegisteredTags[tagno];
if (tag.length()> 1 || !ispunct((unsigned char)tag[0]))
R.m_PossibleWordTags.insert(tagno);
}
}
return R;
}
