//Try to predict double-word surname if the last part is found in dictionary.
//Then check if the first part is found in dictionary as surname.
//If it is not then try predicting it.
bool CWord::PredictHyphenSurname()
{
if (!((m_typ == Hyphen || m_typ == HypDiv || m_typ == DivWord) && m_variant.size() > 0))
return false;
size_t ii = m_txt.find('-');
if (ii == Wtroka::npos)
return false;
//only one hyphen is allowed
if (ii != m_txt.rfind('-'))
return false;
Wtroka strRightPart = m_txt.substr(ii + 1);
if (strRightPart.size() < 1 || !::IsUpper(strRightPart[0]))
return false;
int iH = -1;
THomonymGrammems rightPartGrammems;
Wtroka strRightPartLemma;
if (!RightPartIsSurname(iH, rightPartGrammems, strRightPartLemma))
return false;
ii = m_txt.find('-'); //unnecessary call?
if (ii == Wtroka::npos)
return false;
Wtroka strFirstPart = m_txt.substr(0, ii);
TMorph::ToLower(strFirstPart);
//look in morphology
THomonymVector res;
TMorph::GetDictHomonyms(strFirstPart, res);
bool found = false;
for (size_t i = 0; i < res.size(); ++i) {
if (!found && res[i]->HasGrammem(gSurname) &&
NGleiche::Gleiche(res[i]->Grammems, rightPartGrammems, NGleiche::GenderNumberCaseCheck)) {
found = true;
Wtroka joined_lemma = res[i]->GetLemma() + '-' + strRightPartLemma;
AddHyphenSurnameLemma(iH, rightPartGrammems, joined_lemma);
}
}
if (found)
return true;
//if the word was in morphology then do not do any further predictions
if (res.size() > 0)
return false;
yvector<TSurnamePredictor::TPredictedSurname> out;
TMorph::PredictSurname(strFirstPart, out);
if (out.size() > 0 && NGleiche::Gleiche(out[0].FlexGrammars, rightPartGrammems.Forms(), NGleiche::GenderNumberCaseCheck)) {
Wtroka joined_lemma = out[0].Lemma + '-' + strRightPartLemma;
AddHyphenSurnameLemma(iH, rightPartGrammems, joined_lemma);
return true;
}
return false;
}
bool CWord::RightPartIsSurname(int& iH, THomonymGrammems& grammems, Wtroka& strLemma)
{
iH = HasMorphNounWithGrammems_i(TGramBitSet(gSurname));
if (iH != -1) {
CHomonym& h = GetRusHomonym(iH);
grammems = h.Grammems;
strLemma = h.GetLemma();
size_t ii = strLemma.find('-');
YASSERT(ii != Wtroka::npos);
strLemma = strLemma.substr(ii + 1);
return true;
}
//if this word is in morphology - do not try to predict
if (IsDictionary())
return false;
size_t ii = m_txt.find('-');
if (ii == Wtroka::npos)
return false;
Wtroka strRightPart = m_txt.substr(ii + 1);
TMorph::ToLower(strRightPart);
yvector<TSurnamePredictor::TPredictedSurname> out;
if (!TMorph::PredictSurname(strRightPart, out))
return false;
TGrammarBunch newForms;
NSpike::ToGrammarBunch(out[0].StemGrammar, out[0].FlexGrammars, newForms);
grammems.Reset(newForms);
strLemma = out[0].Lemma;
return true;
}
bool THomonymInflector::FindInForms(const THomonymGrammems& forms, const TGramBitSet& grammems, TGramBitSet& resgram) {
using NInfl::DefaultFeatures;
using NInfl::TFeature;
if (forms.HasForms()) {
for (THomonymGrammems::TFormIter it = forms.IterForms(); it.Ok(); ++it)
if (it->HasAll(grammems)) {
resgram = *it;
return true;
}
} else if (forms.IsIndeclinable() && DefaultFeatures().BitSet(TFeature::Case, TFeature::Number).HasAll(grammems)) {
resgram = grammems;
return true;
}
return false;
}
void CSentence::AddFIOWS(const CFIOOccurence& fioOccurence, const SFullFIO& fio, int iSimilarOccurencesCount)
{
TIntrusivePtr<CFioWordSequence> fioWS(new CFioWordSequence(fio));
*(CFIOOccurence*)(fioWS.Get()) = fioOccurence;
fioWS->PutWSType(FioWS);
if (fio.m_Genders.any()) {
THomonymGrammems gram = fioWS->GetGrammems();
gram.Replace(NSpike::AllGenders, fio.m_Genders);
fioWS->SetGrammems(gram);
}
fioWS->m_iSimilarOccurencesCount = iSimilarOccurencesCount;
bool isManualFio = true;
SWordHomonymNum wh = fioOccurence.m_NameMembers[Surname];
if (wh.IsValid())
if (m_Words.GetWord(wh).IsMultiWord())
isManualFio = false;
if (!fio.m_bFoundSurname &&
fioOccurence.m_NameMembers[Surname].IsValid() &&
!(fioOccurence.m_NameMembers[FirstName].IsValid() ||
fioOccurence.m_NameMembers[InitialName].IsValid()) &&
isManualFio) {
CNameFinder nameFinder(m_Words);
//если не смогли среди предсказанных фамилий
//найти совпадающую с фамилией из fio, то вываливаемся
if (!nameFinder.PredictSingleSurname(*fioWS, fio))
return;
}
fioWS->ClearLemmas();
if (!fio.m_strSurname.empty()) {
Wtroka capLemma;
if (fioOccurence.m_NameMembers[Surname].IsValid()) {
const CWord& w = m_Words.GetWord(fioOccurence.m_NameMembers[Surname]);
capLemma = GetCapitalizedLemma(w, -1, fio.m_strSurname);
} else {
capLemma = fio.m_strSurname;
NStr::ToFirstUpper(capLemma);
}
fioWS->AddLemma(SWordSequenceLemma(fio.m_strSurname, capLemma));
}
if (!fio.m_strName.empty()) {
Wtroka capLemma;
if (fioOccurence.m_NameMembers[FirstName].IsValid()) {
const CWord& w = m_Words.GetWord(fioOccurence.m_NameMembers[FirstName]);
capLemma = GetCapitalizedLemma(w, -1, fio.m_strName);
} else if (fioOccurence.m_NameMembers[InitialName].IsValid()) {
const CWord& w = m_Words.GetWord(fioOccurence.m_NameMembers[InitialName]);
capLemma = GetCapitalizedLemma(w, -1, fio.m_strName);
} else {
capLemma = fio.m_strName;
TMorph::ToTitle(capLemma);
}
fioWS->AddLemma(SWordSequenceLemma(fio.m_strName, capLemma));
}
if (!fio.m_strPatronomyc.empty()) {
Wtroka capLemma;
if (fioOccurence.m_NameMembers[Patronomyc].IsValid()) {
const CWord& w = m_Words.GetWord(fioOccurence.m_NameMembers[Patronomyc]);
capLemma = GetCapitalizedLemma(w, -1, fio.m_strPatronomyc);
} else if (fioOccurence.m_NameMembers[InitialPatronomyc].IsValid()) {
const CWord& w = m_Words.GetWord(fioOccurence.m_NameMembers[InitialPatronomyc]);
capLemma = GetCapitalizedLemma(w, -1, fio.m_strPatronomyc);
} else {
capLemma = fio.m_strPatronomyc;
TMorph::ToTitle(capLemma);
}
fioWS->AddLemma(SWordSequenceLemma(fio.m_strPatronomyc, capLemma));
}
TakeFioWS(fioWS);
}
SWordHomonymNum CMultiWordCreator::AddMultiWordInt(CWordSequence* ws, bool takeOnwership,
const TGramBitSet& newPos, const CWordsPair& searchAreaWP)
{
SWordHomonymNum wh = ws->GetMainWord();
Wtroka stmp;
SWordHomonymNum newWH;
CWord* pNewWord = GetWordForMultiWord(*ws, stmp, newWH);
pNewWord->m_SourceWords.SetPair(ws->FirstWord(), ws->LastWord());
TGramBitSet art_grammems; // output grammems of article
Wtroka article_title;
TKeyWordType article_type = NULL;
if (ws->HasGztArticle()) {
const TGztArticle& gzt_article = ws->GetGztArticle();
article_title = gzt_article.GetTitle();
article_type = gzt_article.GetType();
const NGzt::TMessage* lemma = gzt_article.GetLemmaInfo();
if (lemma != NULL)
art_grammems = gzt_article.GetLemmaOutputGrammems(*lemma);
} else if (ws->HasAuxArticle()) {
const article_t* pArt = GlobalDictsHolder->GetAuxArticle(ws->GetAuxArticleIndex());
art_grammems = pArt->get_new_pos();
article_title = pArt->get_title();
article_type = pArt->get_kw_type();
}
THomonymGrammems newGram;
if (!ws->GetGrammems().Empty()) {
newGram = ws->GetGrammems();
if (!newGram.HasForms() && wh.IsValid())
newGram.SetPOS(m_Words[wh].Grammems.GetPOS());
} else if (wh.IsValid() && HasToAddGrammemsFromMainWord(*ws))
newGram = m_Words[wh].Grammems;
MergeGrammems(newGram, art_grammems, newPos);
THomonymPtr pNewHom;
if (pNewWord->IsMultiWord() && (pNewWord->GetSourcePair().Size() != 1 || !wh.IsValid())) {
newWH.m_HomNum = FindOrMakeMultiwordHomonym(*ws, *pNewWord, article_type, newGram, pNewHom);
YASSERT(newWH.IsValid());
}
if (pNewHom.Get() == NULL) {
if (!pNewWord->IsMultiWord()) {
if (wh.IsValid())
newWH = wh;
else {
// just take the first homonym
newWH.m_bOriginalWord = true;
newWH.m_WordNum = pNewWord->GetSourcePair().FirstWord();
newWH.m_HomNum = pNewWord->IterHomonyms().GetID();
}
}
YASSERT(newWH.IsValid());
//часто бывает ситуация, когда мы вынуждены клонировать абсолютно одинаковые
//омонимы, различающиеся только приписанными статьями из aux_dic,
//в случае с geo_thesaurus.cxx это чревато порождением огромного количества омонимов
//(боле 50 для "Петров"), тогда если статьи не отличаются друг от друга полем СОСТАВ
//приписываемыми граммемами, ЧР и KWType, то мы омонимы не клонируем а дополнительные статьи
//записываем в CHomonym::m_KWtype2Articles. Это происходит в CWord::PutArticleIndex.
//если мы считаем, что найденные статьи для одного и того же омонима ничем не отличаются,
//то главное слово для неотличающихся стаей у ws одно и то же и ему приписана
//первая попавшаяся среди неразличимы статья
//например статьи "_петрова_2" и "_петрова_3" для нас одинаковы (отличаются только ГЕО_ЧАСТЬ
//а это неважно для парсера) и незачем плодить омонимы
bool bCloneAnyway = (!newGram.Empty() && !(m_Words[newWH].Grammems == newGram)) ||
!GlobalDictsHolder->BuiltinKWTypes().IsGeo(article_type);
if (ws->HasAuxArticle())
newWH.m_HomNum = m_Words.GetWord(newWH).PutAuxArticle(newWH.m_HomNum, ws->GetAuxArticleIndex(), bCloneAnyway);
else
newWH.m_HomNum = m_Words.GetWord(newWH).PutGztArticle(newWH.m_HomNum, ws->GetGztArticle(), bCloneAnyway);
}
YASSERT(newWH.IsValid());
AddFoundArticle(article_type, article_title, newWH, searchAreaWP);
CHomonym& h = m_Words[newWH];
h.SetSourceWordSequence(ws);
if (!newGram.Empty())
h.SetGrammems(newGram);
if (takeOnwership) {
if (!ws->HasLemmas())
NormalizeMultiWordHomonym(pNewWord, &h);
m_wordSequences.push_back(ws);
}
return newWH;
}
static void MergeGrammems(THomonymGrammems& dst, const TGramBitSet& art_grammems, const TGramBitSet& newPos)
{
// first, reset Part Of Speech if any
if (newPos.HasAny(TMorph::AllPOS()))
dst.SetPOS(newPos);
else if (art_grammems.HasAny(TMorph::AllPOS()))
dst.SetPOS(art_grammems);
// take other grammems from @art_grammems if any
TGramBitSet other = art_grammems & ~TMorph::AllPOS();
if (other.any()) {
// if there is a form with such grammems - just leave it alone and drop the rest ones
bool found = false;
for (THomonymGrammems::TFormIter it = dst.IterForms(); it.Ok(); ++it)
if (it->HasAll(other)) {
dst.ResetSingleForm(*it);
found = true;
}
if (!found) {
// otherwise merge all forms and replace grammems by classes
TGramBitSet newForm = dst.All();
newForm.ReplaceByMaskIfAny(art_grammems, NSpike::AllCases);
newForm.ReplaceByMaskIfAny(art_grammems, NSpike::AllGenders);
newForm.ReplaceByMaskIfAny(art_grammems, NSpike::AllNumbers);
const TGramBitSet anim(gAnimated, gInanimated);
newForm.ReplaceByMaskIfAny(art_grammems, anim);
newForm.ReplaceByMaskIfAny(art_grammems, NSpike::AllTimes);
newForm.ReplaceByMaskIfAny(art_grammems, NSpike::AllPersons);
// just add the rest non-classified grammems
static const TGramBitSet nonclassified = ~(NSpike::AllCases | NSpike::AllGenders | NSpike::AllNumbers |
anim | NSpike::AllTimes | NSpike::AllPersons);
newForm |= art_grammems & nonclassified;
dst.Reset(newForm);
}
}
// if we still do not known POS, apply some workarounds:
if (dst.GetPOS().none()) {
dst.SetPOS(TGramBitSet(gSubstantive));
if (!dst.HasAny(NSpike::AllCases))
dst.Add(NSpike::AllCases);
if (!dst.HasAny(NSpike::AllGenders))
dst.Add(NSpike::AllGenders);
if (!dst.HasAny(NSpike::AllNumbers))
dst.Add(NSpike::AllNumbers);
}
// set a noun or adj without additional grammem as indeclinable
if (!dst.HasAny(~TMorph::AllPOS()) &&
(art_grammems.Has(gSubstantive) || TMorph::IsFullAdjective(art_grammems)))
dst.Add(NSpike::AllCases | NSpike::AllGenders | NSpike::AllNumbers);
}