敏感词检测

<?php


class WordFilterService
{
    protected $dict;//敏感词字典

    public function __construct()
    {
        $this->loadDataFormFile();
    }

    /**
     * 加载敏感词字典
     *
     *
     */
    protected function loadDataFormFile()
    {
        $arr = $this->getWord();
        //将敏感词加入此次节点
        foreach ($arr as $value) {
            $this->addWords(trim($value));
        }
    }


    /**
     * 获取敏感词库
     *
     * @return array
     */
    public function getWord():array
    {
        return ['大傻瓜', '笨蛋'];
    }



    /**
     * 分割文本
     *
     * @param string $str
     * @return array
     */
    protected function splitStr($str):array
    {
        //将字符串分割成组成它的字符
        // 其中/u 表示按unicode(utf-8)匹配（主要针对多字节比如汉字），否则默认按照ascii码容易出现乱码
        return preg_split("//u", $str, -1, PREG_SPLIT_NO_EMPTY);
    }

    /**
     * 添加敏感字至节点
     *
     * @param string $words
     */
    protected function addWords(string $words)
    {
        //1.分割字典
        $wordArr = $this->splitStr($words);
        $curNode = &$this->dict;

        foreach ($wordArr as $char) {
            if (!isset($curNode)) {
                $curNode[$char] = [];
            }
            $curNode = &$curNode[$char];
        }
        //标记到达当前节点完整路径为"敏感词"
        @$curNode['end']++;
    }

    /**
     * 敏感词校验
     *
     * @param string $str ;需要校验的字符串
     * @param int $level ;屏蔽词校验等级 1-只要顺序包含都屏蔽；2-中间间隔skipDistance个字符就屏蔽；3-全词匹配即屏蔽
     * @param int $skipDistance ;允许敏感词跳过的最大距离，如笨aa蛋a傻瓜等等
     * @param int $handleStyle ;处理方式  1：替换  2：标红
     * @param string $replace ;替换字符
     * @return array
     */
    public function filter(string $str,int $level = 3,int $skipDistance = 2,int $handleStyle = 2,string $replace = '*'):array
    {
        $str = strip_tags($str);//过滤掉html标签
        //允许跳过的最大距离
        if ($level == 1) {
            $maxDistance = strlen($str) + 1;
        } elseif ($level == 2) {
            $maxDistance = max($skipDistance, 0) + 1;
        } else {
            $maxDistance = 2;
        }
        $strArr = $this->splitStr($str);
        $strLength = count($strArr);

//        $isSensitive = false;
        $isMatch = false;

        $match_arr = [];//匹配敏感词
        for ($i = 0; $i < $strLength; $i++) {
            //判断当前敏感字是否有存在对应节点
            $curChar = $strArr[$i];
            if (!isset($this->dict[$curChar])) {
                continue;
            }
//            $isSensitive = true; //引用匹配到的敏感词节点
            $curNode = &$this->dict[$curChar];
            $dist = 0;
            $matchIndex = [$i]; //匹配后续字符串是否match剩余敏感词
            for ($j = $i + 1; $j < $strLength && $dist < $maxDistance; $j++) {

                if (!isset($curNode[$strArr[$j]])) {
                    $dist++;
                    continue;
                }
                //如果匹配到的话，则把对应的字符所在位置存储起来，便于后续敏感词替换
                $matchIndex[] = $j;
                //继续引用
                $curNode = &$curNode[$strArr[$j]];
            }
            //判断是否已经到敏感词字典结尾，是的话，进行敏感词替换
            if (isset($curNode['end'])) {
                $isMatch = true;
                foreach ($matchIndex as $index) {
                    @$match_arr[$i] .= $strArr[$index];
                    if ($handleStyle == 1) {
                        $strArr[$index] = $replace;
                    } elseif ($handleStyle == 2) {
                        $strArr[$index] = '<span style="color:red;">' . $strArr[$index] . '</span>';
                    }
                }
                $i = max($matchIndex);

            }
        }
        $text = implode('', $strArr);

        return ['match' => $isMatch, 'text' => $text, 'match_arr' => $match_arr];

    }


    /**
     * 敏感词检测
     *
     * @param string $str
     * @return array
     */
    public function sensitiveWordFilter(string $str):array
    {
        //敏感词过滤
        $filter_res = $this->filter($str);
        if (isset($filter_res) && $filter_res['match'] == true) {
            $match_arr = array_unique($filter_res['match_arr']);
            $filter_words = '';
            foreach ($match_arr as $v) {
                $filter_words .= '【' . $v . '】';
            }
            return ['code' => 0, 'msg' => '含有敏感词：' . $filter_words . '请修改'];
        }

        return ['code' => 1, 'msg' => '没有敏感词'];
    }
}



$str = '你好呀，大傻瓜';


$obj = new WordFilterService();

$res = $obj->sensitiveWordFilter($str);


var_dump($res);