使用zend Framework的lucene进行全文检索——中文分词

zf本身没有提供中文分词算法，具体应用中要自己写。我这里使用简单的二元分词算法（只在utf-8下工作正常，对于其他字符集，请修改程序）。

第一步、如何测试分词算法的输出。
在zf 的手册中没有提到，我这里简单给个例子：

setInput($value, 'utf-8'); 
         
    $position     = 0; 
        $tokenCounter = 0; 
         while (($token = $analyzer->nextToken()) !== null) { 
            $tokenCounter++; 
            $tokens[] = $token; 
         } 
        print_r($tokens); 
?>

这里使用是zf默认的分词算法Zend_Search_Lucene_Analysis_Analyzer_Common_Text。另外你可以加上一个过滤方法。比如说过滤一些单词，比如“is”,”a “之类的。

_cnStopWords = $cnStopWords; 
     } 

    /** 
      * Reset token stream 
      */ 
    public function reset() 
     { 
        $this->_position = 0; 
        $search = array(",", "/", "\", ".", ";", ":", """, "!", "~", "`", "^", "(", ")", "?", "-", "t", "n", "'", "<", ">", "r", "rn", "$", "&", "%", "#", "@", "+", "=", "{", "}", "[', ']", "：", "）", "（", "．", "。", "，", "！", "；", "“", "”", "‘", "’", "［", "］", "、", "—", "　", "《", "》", "－", "…", "【", "】",); 
        $this->_input = str_replace($search,' ',$this->_input); 
        $this->_input = str_replace($this->_cnStopWords,' ',$this->_input); 
     } 

    /** 
      * Tokenization stream API 
      * Get next token 
      * Returns null at the end of stream 
      * 
      * @return Zend_Search_Lucene_Analysis_Token|null 
      */ 
    public function nextToken() 
     { 
         if ($this->_input === null) { 
             return null; 
         } 
         while ($this->_position < strlen($this->_input)) { 
             while ($this->_position < strlen($this->_input) && 
                    $this->_input[$this->_position]==' ' ) { 
                $this->_position++; 
             } 
            $termStartPosition = $this->_position;       
            $temp_char = $this->_input[$this->_position]; 
            $isCnWord = false; 
             if(ord($temp_char)>127){   
                $i = 0;        
                 while ($this->_position < strlen($this->_input) && 
                ord( $this->_input[$this->_position] )>127) { 
                    $this->_position = $this->_position + 3; 
                    $i ++; 
                     if($i==2){ 
                        $isCnWord = true; 
                         break; 
                     } 
                 } 
                 if($i==1)continue; 
             }else{ 
                 while ($this->_position < strlen($this->_input) && 
                ctype_alnum( $this->_input[$this->_position] )) { 
                    $this->_position++; 
                 } 
             } 
             if ($this->_position == $termStartPosition) { 
                 return null; 
             } 

            $token = new Zend_Search_Lucene_Analysis_Token( 
                                      substr($this->_input, 
                                             $termStartPosition, 
                                             $this->_position - $termStartPosition), 
                                      $termStartPosition, 
                                      $this->_position); 
            $token = $this->normalize($token); 
             if($isCnWord)$this->_position = $this->_position - 3; 
             if ($token !== null) { 
                 return $token; 
             } 
         } 
         return null; 
     } 
     
} 
?>

第二步、自定义自己的分词算法，可以参考手册，或者自己看Zend_Search_Lucene_Analysis_Analyzer_Common_Text类的实现。
其中要注意的是过滤这点。由于我们的分词是二元分词，如果要过滤一些比如“的”、“啊”之类的单词，是无法使用内置的Tokens Filtering。我们需要是分词前先过滤调。这个可以在reset()里面实现
例子。

测试分词输出demo

setCnStopWords($cnStopWords); 
        $analyzer->addFilter($stopWordsFilter); 
        $value = 'this is " a test【中文】的测试'; 
        $analyzer->setInput($value, 'utf-8'); 
         
        $position     = 0; 
        $tokenCounter = 0; 
         while (($token = $analyzer->nextToken()) !== null) { 
            $tokenCounter++; 
            $tokens[] = $token; 
         } 
        print_r($tokens); 
?>