基于信息熵原理分词这个概念很早了，用php实现了个，一气呵成，代码自然是，反正我也没有二次检查，呵呵。不过耗费内存是真的，真的很消耗内存！
写这个的好处就是我明白了很多东西…
代码如下
<?php
/**
* 基于信息熵的无词典分词
*/
class partword{
/**
* 词语最大长度
*/
public $maxwordlen = 5;
/**
* 需要进行分词的文字
*/
public $text;
/**
* 字符串长度
*/
private $len;
/**
* 切分的单个文字
*/
private $textarr;
/**
* 文字信息数组
*/
private $wordinfo;
/**
* 初始化文字，将文字进行分割
*/
public function initWords(){
$i = 0;
$len = strlen($this->text);
$strarr = array();
$word = '';//单词缓存
$en_write = False;
while( $i<$len ){
$ascnum = ord($this->text[$i]);
if( $ascnum >=224 ){ //高位大于244，utf8规定为3个字节
$zh_word = substr($this->text, $i, 3);
$en_write = True;
$i += 3;
}elseif( $ascnum >=192 ){ //高位大于192，utf8规定为2个字节
$zh_word = substr($this->text, $i, 2);
$i+=2;
$en_write = True;
}elseif( $ascnum <= 32 ){
//空字符,英文的空格
$en_write = True;
$i += 1;
}else{
$en_write = False;
$word .= substr($this->text, $i, 1);
$i += 1;
}
if( $en_write && $word != '' ){
$strarr[] = $word;
$word = '';
$en_write = False;
$this->len += 1;
}
if( $zh_word !='' ){
$strarr[] = $zh_word;
$zh_word = '';
$this->len += 1;
}
}
return $this->textarr = $strarr;
}
/**
* 发送文本
*/
public function sendText($text){
$this->text = $text;
$this->initWords();
$this->calInfo();
$this->calEntropy();
}
/**
* 整理词汇，用于后面的凝聚度和信息熵获取
*/
public function calInfo(){
//
if( !isset($this->textarr) ) die("no text");
$newarr = array();
$word = '';
$i = 0;
/*
遍历整个单个文字数组
将所设定的词组阈值之内的所有可能的词进行提取
l=> 左边文字的位置
r=> 右侧文字的位置
c=> 词频
*/
for ($i=0,$len = count($this->textarr); $i<$len; $i++) {
$word = '';
for( $wordlen = 0; $wordlen < $this->maxwordlen; $wordlen++ ){
if( !isset( $this->textarr[$i+$wordlen] ) ) continue;
$word .= $this->textarr[$i+$wordlen];
$newarr[$word]['w'] = $word; //左侧词
$newarr[$word]['l'][] = $i; //左侧词
$newarr[$word]['r'][] = $i+$wordlen; //右侧词
//统计词频
if( !isset($newarr[$word]['c']) ) $newarr[$word]['c'] = 1;
else $newarr[$word]['c'] += 1;
}
}
return $this->wordinfo = $newarr;
}
/**
* 计算信息熵
*/
public function calEntropy(){
foreach ($this->wordinfo as $key => $v) {
$left_entropy = 0;
$right_entropy = 0;
//求左邻字的信息熵,需要大于0才可以
$left_word = array();
//print_r($v);
foreach( $v['l'] as $l ){
if( $l >0 ){
if( !isset($left_word[ $this->textarr[$l-1] ]) )
$left_word[ $this->textarr[$l-1] ] = 1;
else
$left_word[ $this->textarr[$l-1] ] += 1;
}
}
foreach( $left_word as $word=>$num ){
$left_entropy += -($num/$v['c'])*log($num/$v['c']);
}
//求右邻字的信息熵,需要小于整个长度
$right_word = array();
foreach( $v['r'] as $r ){
if( $r < $this->len-1 ){
if( !isset($right_word[ $this->textarr[$r+1] ]) )
$right_word[ $this->textarr[$r+1] ] = 1;
else
$right_word[ $this->textarr[$r+1] ] += 1;
}
}
foreach( $right_word as $word=>$num ){
$right_entropy += -($num/$v['c'])*log($num/$v['c']);
}
//echo $key."--".$left_entropy."--".$right_entropy."\n";
//取坐信息熵和又信息熵里面最小的一个
$entropy = ($left_entropy < $right_entropy) ? $left_entropy : $right_entropy;
$this->wordinfo[$key]['entropy'] = $entropy;
unset($this->wordinfo[$key]['l']);
unset($this->wordinfo[$key]['r']);
}
unset($this->textarr);
return $this->wordinfo;
}
/**
* 对词进行排序,根据信息熵排序
*/
public function sortResult(){
usort($this->wordinfo, array(&$this,'sortByEntropy'));
return $this->wordinfo;
}
private function sortByEntropy($a,$b){
if( $a['entropy'] === $b['entropy'] ) return 0;
return ($a['entropy'] > $b['entropy']) ? -1: 1;
}
}
//应用
header("Content-type:text/html;charset=utf-8");
$smen = memory_get_usage();
$partword = new partword();
$text = "吃葡萄不吐葡萄皮不吃葡萄倒吐葡萄皮";
$partword->sendText($text);
print_r( $partword->sortResult() );
$emen = memory_get_usage();
echo (($emen - $smen)/1024) ."kb";
?>
赞赏
微信赞赏
支付宝赞赏
