bool TBagOfWords::Update(const TStrV& TokenStrV) {
// Generate Ngrams if necessary
TStrV NgramStrV;
GenerateNgrams(TokenStrV, NgramStrV);
// process tokens to update DF counts
bool UpdateP = false;
if (IsHashing()) {
// consolidate tokens and get their hashed IDs
TIntSet TokenIdH;
for (int TokenStrN = 0; TokenStrN < NgramStrV.Len(); TokenStrN++) {
const TStr& TokenStr = NgramStrV[TokenStrN];
TInt TokenId = TokenStr.GetHashTrick() % HashDim;
TokenIdH.AddKey(TokenId);
if (IsStoreHashWords()) { HashWordV[TokenId].AddKey(TokenStr); }
}
// update document counts
int KeyId = TokenIdH.FFirstKeyId();
while (TokenIdH.FNextKeyId(KeyId)) {
const int TokenId = TokenIdH.GetKey(KeyId);
// update DF
DocFqV[TokenId]++;
}
} else {
// consolidate tokens
TStrH TokenStrH;
for (int TokenStrN = 0; TokenStrN < NgramStrV.Len(); TokenStrN++) {
const TStr& TokenStr = NgramStrV[TokenStrN];
TokenStrH.AddKey(TokenStr);
}
// update document counts and update vocabulary with new tokens
int KeyId = TokenStrH.FFirstKeyId();
while (TokenStrH.FNextKeyId(KeyId)) {
// get token
const TStr& TokenStr = TokenStrH.GetKey(KeyId);
// different processing for hashing
int TokenId = TokenSet.GetKeyId(TokenStr);
if (TokenId == -1) {
// new token, remember the dimensionality change
UpdateP = true;
// remember the new token
TokenId = TokenSet.AddKey(TokenStr);
// increase document count table
const int TokenDfId = DocFqV.Add(0);
// increase also the old count table
OldDocFqV.Add(0.0);
// make sure we DF vector and TokenSet still in sync
IAssert(TokenId == TokenDfId);
IAssert(DocFqV.Len() == OldDocFqV.Len());
}
// document count update
DocFqV[TokenId]++;
}
}
// update document count
Docs++;
// tell if dimension changed
return UpdateP;
}
TStr TStrUtil::GetStr(const TStrH& StrH, const TStr& FieldDelimiterStr, const TStr& DelimiterStr) {
if (StrH.Empty()) {return TStr();}
TChA ResChA;
int KeyId = StrH.FFirstKeyId();
while (StrH.FNextKeyId(KeyId)) {
if (!ResChA.Empty()) { ResChA+=DelimiterStr; }
ResChA+=StrH.GetKey(KeyId);
ResChA+=FieldDelimiterStr;
ResChA+=StrH[KeyId].GetStr();
}
return ResChA;
}
void TFtrGenToken::Update(const TStr& Val) {
TStrV TokenStrV; GetTokenV(Val, TokenStrV); TStrH TokenStrH;
for (int TokenStrN = 0; TokenStrN < TokenStrV.Len(); TokenStrN++) {
const TStr& TokenStr = TokenStrV[TokenStrN];
TokenStrH.AddKey(TokenStr);
}
int KeyId = TokenStrH.FFirstKeyId();
while (TokenStrH.FNextKeyId(KeyId)) {
const TStr& TokenStr = TokenStrH.GetKey(KeyId);
TokenH.AddDat(TokenStr)++;
}
Docs++;
}
void TCycBs::GetRelNmV(TStrV& RelNmV){
TStrH RelNmH;
// traverse vertices
for (int VId=0; VId<GetVIds(); VId++){
TStr VNm=GetVNm(VId);
TCycVrtx& Vrtx=GetVrtx(VId);
// traverse edges per vertice
for (int EdgeN=0; EdgeN<Vrtx.GetEdges(); EdgeN++){
// extract relation name
TCycEdge& Edge=Vrtx.GetEdge(EdgeN);
TStr RelNm=GetVNm(Edge.GetRelId());
// add relation name to the pool
RelNmH.AddKey(RelNm);
}
}
// extract relation-name-vector
RelNmH.GetKeyV(RelNmV);
}
void TCpDoc::SaveAsfaToCpd(const TStr& InFPath, const TStr& OutCpdFNm){
// create output file
PSOut SOut=TFOut::New(OutCpdFNm);
// traverse files
TStrH AccessionIdH;
TFFile FFile(TStr::GetNrFPath(InFPath)+"*.Asfa"); TStr AsfaFNm;
while (FFile.Next(AsfaFNm)){
printf("Processing file '%s'\n", AsfaFNm.CStr());
PSIn SIn=TFIn::New(AsfaFNm);
TILx Lx(SIn, TFSet(iloRetEoln, iloExcept));
Lx.GetSym(syLn, syEof);
while (Lx.Sym!=syEof){
// Query Line
TStr QueryLnStr=Lx.Str;
TStrV QueryStrV; QueryLnStr.SplitOnAllCh('\t', QueryStrV, false);
IAssert(QueryStrV[0]=="Query");
// RecordNo Line
Lx.GetSym(syLn); TStr RecNoLnStr=Lx.Str;
TStrV RecNoStrV; RecNoLnStr.SplitOnAllCh('\t', RecNoStrV, false);
IAssert(RecNoStrV[0]=="RecordNo");
//int RecN=RecNoStrV[1].GetInt();
// fields (format: Short-Name Tab Long-Name Tab Value-String)
TStr TitleStr, AbstractStr, PublicationYearStr, AccessionId;
TStrV AuthorNmV; TStrV TermNmV1, TermNmV2;
while (true){
Lx.GetSym(syLn); TStr FldLnStr=Lx.Str;
TStrV FldStrV; FldLnStr.SplitOnAllCh('\t', FldStrV, false);
if (FldStrV[0]=="----"){
if (!AccessionIdH.IsKey(AccessionId)){
AccessionIdH.AddKey(AccessionId);
// create & save cpd document
PCpDoc CpDoc=TCpDoc::New();
CpDoc->DocNm=AccessionId;
CpDoc->DateStr=PublicationYearStr;
CpDoc->TitleStr=TitleStr;
CpDoc->ParStrV.Add(AbstractStr);
CpDoc->TopCdNmV=TermNmV1;
CpDoc->GeoCdNmV=TermNmV2;
CpDoc->IndCdNmV=AuthorNmV;
CpDoc->Save(*SOut);
} else {/*printf("[%s]", AccessionId.CStr());*/}
break;
} else
if (FldStrV[0]=="TI"){
TitleStr=FldStrV[2];
} else if (FldStrV[0]=="TI"){
TitleStr=FldStrV[2];
} else if (FldStrV[0]=="AU"){
FldStrV[2].SplitOnAllCh(';', AuthorNmV);
for (int StrN=0; StrN<AuthorNmV.Len(); StrN++){AuthorNmV[StrN].ToTrunc();}
} else if (FldStrV[0]=="AB"){
AbstractStr=FldStrV[2];
} else if (FldStrV[0]=="PY"){
PublicationYearStr=FldStrV[2];
} else if (FldStrV[0]=="DE"){
FldStrV[2].SplitOnAllCh(';', TermNmV1);
for (int StrN=0; StrN<TermNmV1.Len(); StrN++){TermNmV1[StrN].ToTrunc();}
} else if (FldStrV[0]=="CL"){
FldStrV[2].SplitOnAllCh(';', TermNmV2);
for (int StrN=0; StrN<TermNmV2.Len(); StrN++){TermNmV2[StrN].ToTrunc();}
} else if (FldStrV[0]=="AN"){
AccessionId=FldStrV[2];
}
}
printf("%d\r", AccessionIdH.Len());
Lx.GetSym(syLn, syEof);
}
}
}
