C++处理reuters21578（二）

作者finallyliuyu 出处博客园

 通过C++处理reuters21578（一）的代码，初步形成了两张表单存放训练语料库和测试语料库，由于这两个语料库的个别类别不一致，所以要找到这两个语料库类别的交集，然后最终形成文本分类的训练语料库和测试语料库。以下主函数中完成此功能。

 

 

class  GT_clss
 {
  public :
     GT_clss( string   & s):comparepart(s){}
      bool   operator ()( const   string   & elem)
     {
          return  elem == comparepart;

     }
  private :
      string  comparepart;
 };

 

 

 

数据库中共有多少个类别

vector < string > GetLabels( string  tablename)
 {   vector < string > labels;
  char   *  selectbySpecificId = new   char  [ 1000 ];
 memset(selectbySpecificId, 0 , 1000 );
 sprintf_s(selectbySpecificId, 1000 , " select Categorization from %s  " ,tablename.c_str());
 CoInitialize(NULL);
 _ConnectionPtr pConn(__uuidof(Connection));
 _RecordsetPtr pRst(__uuidof(Recordset));
 pConn -> ConnectionString = " Provider=SQLOLEDB.1;Password=finally;Persist Security Info=True; User ID=sa;Initial Catalog=NewsInfo " ;
 pConn -> Open( "" , "" , "" ,adConnectUnspecified);
 pRst = pConn -> Execute(selectbySpecificId,NULL,adCmdText);
  while ( ! pRst -> rsEOF)
 {
      string  label = (_bstr_t)pRst -> GetCollect( " Categorization " );
      if  ( ! count_if(labels.begin(),labels.end(),GT_clss(label)))
     {
         labels.push_back(label);
     }

     pRst -> MoveNext();

 }
 pRst -> Close();
 pConn -> Close();
 pRst.Release();
 pConn.Release();
 CoUninitialize();
 delete []selectbySpecificId;

  return  labels;



 }

 

 

 

主函数

int  _tmain( int  argc, _TCHAR *  argv[])
{
      int  end;
     // set<string>labels;
    vector < string > labelsTrain = GetLabels( " ReteursTrain " );
    vector < string > labelsTest = GetLabels( " ReteursTest " );
    vector < string > finalLabels;
     for  (vector < string > ::iterator it = labelsTrain.begin();it != labelsTrain.end();it ++ )
    {
        trim( * it, "   " );
    }
     for (vector < string > ::iterator it = labelsTest.begin();it != labelsTest.end();it ++ )
    {
        trim( * it, "   " );

    }
    
     for  (vector < string > ::iterator it = labelsTrain.begin();it != labelsTrain.end();it ++ )
    {
         if  (count_if(labelsTest.begin(),labelsTest.end(),GT_clss( * it)))
        {
            finalLabels.push_back( * it);
        }
    }

     char   *  selectbySpecificId = new   char  [ 1000 ];
    memset(selectbySpecificId, 0 , 1000 );
    sprintf_s(selectbySpecificId, 1000 , " select CArticleName,CAbstract,Categorization from ReteursTest " );
    CoInitialize(NULL);
    _ConnectionPtr pConn(__uuidof(Connection));
    _RecordsetPtr pRst(__uuidof(Recordset));
    _ConnectionPtr pConn2(__uuidof(Connection));
    pConn -> ConnectionString = " Provider=SQLOLEDB.1;Password=finally;Persist Security Info=True; User ID=sa;Initial Catalog=NewsInfo " ;
    pConn2 -> ConnectionString = " Provider=SQLOLEDB.1;Password=finally;Persist Security Info=True; User ID=sa;Initial Catalog=FinallyCorpus " ;
    pConn -> Open( "" , "" , "" ,adConnectUnspecified);
    pConn2 -> Open( "" , "" , "" ,adConnectUnspecified);
    pRst = pConn -> Execute(selectbySpecificId,NULL,adCmdText);
     while ( ! pRst -> rsEOF)
    {
         string  label = (_bstr_t)pRst -> GetCollect( " Categorization " );
        trim(label, "   " );

         if  (count_if(finalLabels.begin(),finalLabels.end(),GT_clss(label)))
        {
             string  ArticleTitle = (_bstr_t)pRst -> GetCollect( " CArticleName " );
             string  ArticleText = (_bstr_t)pRst -> GetCollect( " CAbstract " );
            ArticleTitle = ProcessforMSSQL(ArticleTitle);
            ArticleText = ProcessforMSSQL(ArticleText);
             char   * sqlInsert = new   char [ 1000000 ];
            _variant_t RecordsAffected;
            memset(sqlInsert, 0 , 1000000 );
            sprintf_s(sqlInsert, 1000000 , " insert into ReteursTestingCorpus(CArticleName,CAbstract,Categorization) values('%s','%s','%s') " ,ArticleTitle.c_str(),ArticleText.c_str(),label.c_str());
            pConn2 -> Execute(sqlInsert, & RecordsAffected, - 1 );
            delete []sqlInsert;

            


            
        }
        

        pRst -> MoveNext();

    }
    pRst -> Close();
    pConn -> Close();
    pRst.Release();
    pConn.Release();
    pConn2 -> Close();
    pConn2.Release();
    CoUninitialize();
    delete []selectbySpecificId;

    
    cout << " 两标签集交集为 " << endl;

    cout << finalLabels.size() << endl;

     // DictionaryToDataBase();
    
     // FindFile(L"E:\\新闻语料\\reuters21578");
    

     // pRst=pConn->Execute(,NULL,adCmdText);


    
   cout << " finish " << endl;
    
    
    cin >> end;






}