ICTCLAS分词系统研究（九）--对最终结果做优化调整

在研究（八）中，我们得到了最终的分词结果了，好兴奋呀。不过，还有临门一脚不能忘了，对一些特殊情况做处理。主要是对叠词（相邻的两个字或词一样）及个别词性进行合并处理。

比如，以“一片片的白云很好看”，他的最终分词结果是：

 经过优化后的分词结果：

序号	分词结果
0	一/m 片/q 片/q 的/uj 白云/n 很/d 好看/a

很显然，“一片片”应该为一个整体，没有拆分的必要，看源代码的调整过程：

//Adjust the result with some rules
bool CResult::Adjust(PWORD_RESULT pItem,PWORD_RESULT pItemRet)
...{
    int i=0,j=0;
    unsigned int nLen;
    char sSurName[10],sSurName2[10],sGivenName[10];
    bool bProcessed=false;//Have been processed
    while(pItem[i].sWord[0]!=0)
    ...{
        nLen=strlen(pItem[i].sWord);
        bProcessed=false;
        
        //Rule1: adjust person name
        if(pItem[i].nHandle==28274&&ChineseNameSplit(pItem[i].sWord,sSurName,sSurName2,sGivenName,m_uPerson.m_dict)&&strcmp(pItem[i].sWord,"叶利钦")!=0)//'nr'
        ...{//Divide name into surname and given name
            
            if(sSurName[0])
            ...{
                strcpy(pItemRet[j].sWord,sSurName);
                pItemRet[j++].nHandle=28274;
            }
            if(sSurName2[0])
            ...{
                strcpy(pItemRet[j].sWord,sSurName2);
                pItemRet[j++].nHandle=28274;
            }
            if(sGivenName[0])
            ...{
                strcpy(pItemRet[j].sWord,sGivenName);
                pItemRet[j++].nHandle=28274;
            }
            bProcessed=true;
        }
        //Rule2 for overlap words ABB 一段段、一片片
        else if(pItem[i].nHandle==27904&&strlen(pItem[i+1].sWord)==2&&strcmp(pItem[i+1].sWord,pItem[i+2].sWord)==0)
        ...{//(pItem[i+1].nHandle/256=='q'||pItem[i+1].nHandle/256=='a')&&
            strcpy(pItemRet[j].sWord,pItem[i].sWord);
            strcat(pItemRet[j].sWord,pItem[i+1].sWord);
            strcat(pItemRet[j].sWord,pItem[i+2].sWord);
            pItemRet[j].nHandle=27904;
            j+=1;
            i+=2;
            bProcessed=true;
        }
        //Rule3 for overlap words AA
        else if(nLen==2&&strcmp(pItem[i].sWord,pItem[i+1].sWord)==0)
        ...{
            strcpy(pItemRet[j].sWord,pItem[i].sWord);
            strcat(pItemRet[j].sWord,pItem[i+1].sWord);
             //24832=='a'*256
            pItemRet[j].nHandle=24832;//a
            if(pItem[i].nHandle/256=='v'||pItem[i+1].nHandle/256=='v')//30208='v'8256
            ...{
                pItemRet[j].nHandle=30208;
            }
            if(pItem[i].nHandle/256=='n'||pItem[i+1].nHandle/256=='n')//30208='v'8256
            ...{
                pItemRet[j].nHandle='n'*256;
            }            
            i+=1;
            if(strlen(pItem[i+1].sWord)==2)
            ...{//AAB:洗/洗/脸、蒙蒙亮
                if((pItemRet[j].nHandle==30208&&pItem[i+1].nHandle/256=='n')||
                   (pItemRet[j].nHandle==24832&&pItem[i+1].nHandle/256=='a')
                   )
                ...{
                    strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                    i+=1;
                }
            }
            j+=1;
            bProcessed=true;
        }

        //Rule 4: AAB 洗/洗澡
        else if(nLen==2&&strncmp(pItem[i].sWord,pItem[i+1].sWord,2)==0&&strlen(pItem[i+1].sWord)==4&&(pItem[i].nHandle/256=='v'||pItem[i].nHandle==24832))//v,a
        ...{
            strcpy(pItemRet[j].sWord,pItem[i].sWord);
            strcat(pItemRet[j].sWord,pItem[i+1].sWord);
             //24832=='a'*256
            pItemRet[j].nHandle=24832;//'a'
            if(pItem[i].nHandle/256=='v'||pItem[i+1].nHandle/256=='v')//30208='v'8256
            ...{
                pItemRet[j].nHandle=30208;
            }

            i+=1;
            j+=1;
            bProcessed=true;
        }
        else if(pItem[i].nHandle/256=='u'&&pItem[i].nHandle%256)//uj,ud,uv,uz,ul,ug->u
            pItem[i].nHandle='u'*256;
        else if(nLen==2&&strncmp(pItem[i].sWord,pItem[i+1].sWord,2)==0&&strlen(pItem[i+1].sWord)==4&&strncmp(pItem[i+1].sWord+2,pItem[i+2].sWord,2)==0)
        ...{//AABB 朴朴素素 枝枝叶叶
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                strcat(pItemRet[j].sWord,pItem[i+2].sWord);
                pItemRet[j].nHandle=pItem[i+1].nHandle;
                i+=2;
                j+=1;
                bProcessed=true;
        }
        else if(pItem[i].nHandle==28275)//PostFix
        ...{
            if(m_uPlace.m_dict.IsExist(pItem[i+1].sWord,4))
            ...{
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                pItemRet[j].nHandle=28275;
                i+=1;
                j+=1;
                bProcessed=true;
            }
            else if(strlen(pItem[i+1].sWord)==2&&CC_Find("队",pItem[i+1].sWord))
            ...{
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                pItemRet[j].nHandle=28276;
                i+=1;
                j+=1;
                bProcessed=true;
            }
            else if(strlen(pItem[i+1].sWord)==2&&CC_Find("语文字杯",pItem[i+1].sWord))
            ...{
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                pItemRet[j].nHandle=28282;
                i+=1;
                j+=1;
                bProcessed=true;
            }
            else if(strlen(pItem[i+1].sWord)==2&&CC_Find("裔",pItem[i+1].sWord))
            ...{
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                pItemRet[j].nHandle=28160;
                i+=1;
                j+=1;
                bProcessed=true;
            }
        }
        else if(pItem[i].nHandle==30208||pItem[i].nHandle==28160)//v
        ...{
            if(strlen(pItem[i+1].sWord)==2&&CC_Find("员",pItem[i+1].sWord))
            ...{
                strcpy(pItemRet[j].sWord,pItem[i].sWord);
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                pItemRet[j].nHandle=28160;
                i+=1;
                j+=1;
                bProcessed=true;
            }
        }
        else if(pItem[i].nHandle==28280)
        ...{//www/nx ./w sina/nx; ＥＩＭ/nx  -６０１/m 
            strcpy(pItemRet[j].sWord,pItem[i].sWord);
            pItemRet[j].nHandle=28280;
            while(pItem[i+1].nHandle==28280||strstr(".．",pItem[i+1].sWord)||(pItem[i+1].nHandle==27904&&IsAllNum((unsigned char *)pItem[i+1].sWord)))
            ...{
                strcat(pItemRet[j].sWord,pItem[i+1].sWord);
                i+=1;
            }
            j+=1;
            bProcessed=true;
        }

        if(!bProcessed)
        ...{//If not processed,that's mean: not need to adjust;
         //just copy to the final result
            strcpy(pItemRet[j].sWord,pItem[i].sWord);
            pItemRet[j++].nHandle=pItem[i].nHandle;
        }
        i++;
    }
    pItemRet[j].sWord[0]=0;//Set ending
    return true;
}

 

调整合并后的最终分词结果为：

一片片/m 的/u 白云/n 很/d 好看/a

至此，我们终于完成了整个分词的流程，得到了正确的切分结果^_^