mint/api-v13/app/Tools/TurboSplit.php

<?php

namespace App\Tools;

require_once __DIR__.'/../../public/app/public/casesuf.inc';

use App\Facades\RedisCache;
use App\Models\UserDict;
use App\Models\WordPart;
use Illuminate\Support\Facades\Log;

class TurboSplit
{
    protected $options = [
        'express' => false,
        'c_threshhold' => 0.8,
        'w_threshhold' => 0.8,
        'forward' => true,
        'sandhi_advance' => false,
        'lookup_declension' => true,
        'timeout' => 1000, // 超时放弃
        /**快速查字典-不去尾 */
    ];

    protected $started_at = null;

    protected $node = [];

    protected $path = [];

    protected $isDebug = false;

    // 当前搜索路径信心指数，如果过低，马上终止这个路径的搜索
    protected $currPathCf;

    // 结果数组
    protected $result = [];

    // 过程中最大结果数量
    protected $MAX_RESULT = 100;

    // 返回值最大结果数量
    protected $MAX_RESULT2 = 8;

    // 最大递归深度
    protected $MAX_DEEP = 16;

    // 连音规则表
    protected $sandhi = [
        ['a' => '', 'b' => '', 'c' => '', 'len' => 0, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'a', 'b' => 'a', 'c' => 'ā', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'ā', 'b' => 'ā', 'c' => 'ā', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'ā', 'c' => 'ā', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ā', 'b' => 'a', 'c' => 'ā', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'e', 'c' => 'e', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'i', 'c' => 'i', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'o', 'c' => 'o', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'u', 'c' => 'o', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'a', 'c' => 'o', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'u', 'c' => 'ū', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'u', 'c' => 'u', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'ī', 'c' => 'ī', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'ū', 'c' => 'ū', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'i', 'c' => 'e', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'e', 'b' => 'a', 'c' => 'e', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'i', 'c' => 'ī', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'e', 'c' => 'e', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'a', 'c' => 'ya', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'atth', 'c' => 'atth', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'taṃ', 'b' => 'n', 'c' => 'tann', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'api', 'c' => 'mpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'eva', 'c' => 'meva', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[o]', 'b' => 'iva', 'c' => 'ova', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'o', 'b' => 'a', 'c' => 'o', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'ādi', 'c' => 'ādi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a[ānaṃ]', 'b' => 'a', 'c' => 'ānama', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'iti', 'c' => 'āti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'ca', 'c' => 'ñca', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'iti', 'c' => 'nti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'a', 'c' => 'ma', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ṃ', 'b' => 'a', 'c' => 'm', 'len' => 1, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => '[ṃ]', 'b' => 'ā', 'c' => 'mā', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => 'ṃ', 'b' => 'ā', 'c' => 'mā', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => '[ṃ]', 'b' => 'u', 'c' => 'mu', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => '[ṃ]', 'b' => 'h', 'c' => 'ñh', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => 'ā', 'b' => '[ṃ]', 'c' => 'am', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => 'a', 'b' => '[ṃ]', 'c' => 'am', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => 'ī', 'b' => '[ṃ]', 'c' => 'im', 'len' => 2, 'adj_len' => 0, 'advance' => false, 'cf' => 0.8],
        ['a' => 'ati', 'b' => 'tabba', 'c' => 'atabba', 'len' => 6, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ati', 'b' => 'tabba', 'c' => 'itabba', 'len' => 6, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'iti', 'b' => 'a', 'c' => 'icca', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'uṃ', 'b' => 'a', 'c' => 'uma', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u[ūnaṃ]', 'b' => 'a', 'c' => 'ūnama', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ī[īnaṃ]', 'b' => 'a', 'c' => 'īnama', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'su', 'b' => 'a', 'c' => 'sva', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ā', 'b' => 'iti', 'c' => 'āti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'iti', 'c' => 'āti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'e', 'b' => 'iti', 'c' => 'eti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'ī', 'b' => 'iti', 'c' => 'īti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'iti', 'c' => 'īti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'o', 'b' => 'iti', 'c' => 'oti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'ū', 'b' => 'iti', 'c' => 'ūti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'iti', 'c' => 'ūti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'ṃ', 'b' => 'iti', 'c' => 'nti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'ṃ', 'b' => 'ca', 'c' => 'ñca', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'ṃ', 'b' => 'cāti', 'c' => 'ñcāti', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'ṃ', 'b' => 'cet', 'c' => 'ñcet', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ṃ', 'b' => 'ev', 'c' => 'mev', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.99999],
        ['a' => 'a', 'b' => 'a', 'c' => 'a', 'len' => 1, 'adj_len' => -1, 'advance' => true, 'cf' => 0.99],
        ['a' => 'ī', 'b' => '', 'c' => 'i', 'len' => 1, 'adj_len' => 0, 'advance' => true, 'cf' => 0.9],
    ];

    protected $sandhi2 = [
        ['a' => 'ṃ', 'b' => 'ca', 'c' => 'ñca', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'ṃ', 'b' => 'hi', 'c' => 'ñhi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'ena', 'b' => 'iti', 'c' => 'enāti', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 1.0],
        ['a' => 'a', 'b' => 'iti', 'c' => 'āti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'ā', 'b' => 'iti', 'c' => 'āti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.6],
        ['a' => 'e', 'b' => 'iti', 'c' => 'eti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'i', 'b' => 'iti', 'c' => 'īti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'o', 'b' => 'iti', 'c' => 'oti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'u', 'b' => 'iti', 'c' => 'ūti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'ṃ', 'b' => 'iti', 'c' => 'nti', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'ī', 'b' => 'eva', 'c' => 'iyeva', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9],
        ['a' => 'ī', 'b' => 'eva', 'c' => 'īyeva', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'eva', 'c' => 'uyeva', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ṃ', 'b' => 'eva', 'c' => 'ṃyeva', 'len' => 5, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'eva', 'c' => 'yeva', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'o', 'b' => 'eva', 'c' => 'ova', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ṃ', 'b' => 'eva', 'c' => 'meva', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'eva', 'c' => 'veva', 'len' => 4, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'eva', 'c' => 'eva', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'e', 'b' => 'eva', 'c' => 'eva', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'a', 'b' => 'api', 'c' => 'āpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ā', 'b' => 'api', 'c' => 'āpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'e', 'b' => 'api', 'c' => 'epi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'i', 'b' => 'api', 'c' => 'īpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ī', 'b' => 'api', 'c' => 'īpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'o', 'b' => 'api', 'c' => 'opi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'api', 'c' => 'ūpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ū', 'b' => 'api', 'c' => 'ūpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'u', 'b' => 'api', 'c' => 'upi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
        ['a' => 'ṃ', 'b' => 'api', 'c' => 'mpi', 'len' => 3, 'adj_len' => 0, 'advance' => false, 'cf' => 0.9999],
    ];

    /**
     * Create a new class instance.
     *
     * @return void
     */
    public function __construct($options = [])
    {
        for ($i = 0; $i < $this->MAX_DEEP; $i++) {
            array_push($this->path, ['', 0]);
        }
        foreach ($options as $key => $value) {
            $this->options[$key] = $value;
        }

    }

    /**
     * 从双元音处切开
     *
     * @param  string  $word
     * @return array
     */
    public function splitDiphthong($word)
    {
        // diphthong table双元音表
        $search = ['aa', 'ae', 'ai', 'ao', 'au', 'aā', 'aī', 'aū', 'ea', 'ee', 'ei', 'eo', 'eu', 'eā', 'eī', 'eū', 'ia', 'ie', 'ii', 'io', 'iu', 'iā', 'iī', 'iū', 'oa', 'oe', 'oi', 'oo', 'ou', 'oā', 'oī', 'oū', 'ua', 'ue', 'ui', 'uo', 'uu', 'uā', 'uī', 'uū', 'āa', 'āe', 'āi', 'āo', 'āu', 'āā', 'āī', 'āū', 'īa', 'īe', 'īi', 'īo', 'īu', 'īā', 'īī', 'īū', 'ūa', 'ūe', 'ūi', 'ūo', 'ūu', 'ūā', 'ūī', 'ūū'];
        $replace = ['a-a', 'a-e', 'a-i', 'a-o', 'a-u', 'a-ā', 'a-ī', 'a-ū', 'e-a', 'e-e', 'e-i', 'e-o', 'e-u', 'e-ā', 'e-ī', 'e-ū', 'i-a', 'i-e', 'i-i', 'i-o', 'i-u', 'i-ā', 'i-ī', 'i-ū', 'o-a', 'o-e', 'o-i', 'o-o', 'o-u', 'o-ā', 'o-ī', 'o-ū', 'u-a', 'u-e', 'u-i', 'u-o', 'u-u', 'u-ā', 'u-ī', 'u-ū', 'ā-a', 'ā-e', 'ā-i', 'ā-o', 'ā-u', 'ā-ā', 'ā-ī', 'ā-ū', 'ī-a', 'ī-e', 'ī-i', 'ī-o', 'ī-u', 'ī-ā', 'ī-ī', 'ī-ū', 'ū-a', 'ū-e', 'ū-i', 'ū-o', 'ū-u', 'ū-ā', 'ū-ī', 'ū-ū'];
        // 将双元音拆开
        // step 1 : split at diphthong . ~aa~ -> ~a-a~
        $word1 = str_replace($search, $replace, $word);
        // 按连字符拆开处理
        $arrword = str_getcsv($word1, '-');

        return $arrword;
    }

    /**
     * 查询单词是否存在于词干表
     * 如果存在，返回单词权重
     * 如果不存在，变格后返回权重和语尾长度
     *
     * @param  string  $word
     * @return array(int $wordWeight, int $endingLength)
     */
    public function dict_lookup($word)
    {
        global $case;        // 语尾表
        if (strlen($word) <= 1) {
            return [0, 0];
        }
        $search = $word;

        // 获取单词权重
        $weight = RedisCache::remember(
            'palicanon/wordpart/weight/'.$search,
            config('mint.cache.expire'),
            function () use ($search) {
                return WordPart::where('word', $search)->value('weight');
            }
        );
        if ($weight) {
            // 找到
            $this->log($search.'='.$weight);

            return [$weight, 0];
        } else {
            // 没找到
            if ($this->options['lookup_declension']) {
                return [0, 0];
            }
            // 去除尾查
            $newWord = [];
            for ($row = 0; $row < count($case); $row++) {
                $len = mb_strlen($case[$row][1], 'UTF-8');
                $end = mb_substr($search, 0 - $len, null, 'UTF-8');
                if ($end == $case[$row][1]) {
                    $base = mb_substr($search, 0, mb_strlen($search, 'UTF-8') - $len, 'UTF-8').$case[$row][0];
                    if ($base != $search) {
                        $newWord[$base] = mb_strlen($case[$row][1], 'UTF-8');
                    }
                }
            }
            // 找到权重最高的base
            $base_weight = 0;
            $len = 0;
            foreach ($newWord as $x => $x_len) {
                $weight = RedisCache::remember(
                    'palicanon/wordpart/weight/'.$x,
                    config('mint.cache.expire'),
                    function () use ($x) {
                        return WordPart::where('word', $x)->value('weight');
                    }
                );
                if ($weight) {
                    if ($weight > $base_weight) {
                        $base_weight = $weight;
                        $len = $x_len;
                    }
                }
            }

            return [$base_weight, $len];
        }
    }

    /**
     * 查找某个单词是否在现有词典出现
     * 返回信心指数
     * look up single word in dictionary vocabulary
     * return the confidence value
     */
    public function isExist($word, $adj_len = 0)
    {
        $this->log("正在查询：{$word}");

        $isFound = false;
        $count = 0;
        $wordPart = RedisCache::remember(
            "turbosplit/part/{$word}",
            config('mint.cache.expire'),
            function () use ($word) {
                return implode(',', $this->dict_lookup($word));
            }
        );
        $arrWordPart = explode(',', $wordPart);
        $word_count = $arrWordPart[0];
        if (isset($arrWordPart[1])) {
            $case_len = $arrWordPart[1];
        } else {
            $case_len = 0;
            Log::error('wordPart error value='.$wordPart);
        }

        if ($word_count > 0) {
            $this->log("查到：{$word}:{$word_count}个");
            $isFound = true;
            $count = $word_count + 1;
        }

        // fomular of confidence value 信心值计算公式
        if ($isFound) {
            // 词本身就是语尾时 $len = 0，pow(-0.5, 1.2) 得 NAN；JSON 存不了 NAN，这种值不进缓存
            $cacheKey = 'turbosplit/confidence/'.$word;
            $cf = RedisCache::get($cacheKey);
            if ($cf === null) {
                $len = mb_strlen($word, 'UTF-8') - $case_len;
                $len_correct = 1.2;
                $count2 = 1.1 + pow($count, 1.18);
                $conf_num = pow(1 / $count2, pow(($len - 0.5), $len_correct));
                $cf = round(1 / (1 + 640 * $conf_num), 9);

                if (is_finite($cf)) {
                    RedisCache::put($cacheKey, $cf, config('mint.cache.expire'));
                }
            }

            return $cf;
        } else {
            return -1;
        }
    }

    /**
     * 判断是否超时
     *
     * @return bool
     */
    private function isTimeOut()
    {
        if ($this->started_at) {
            $time = time() - $this->started_at;
            if ($time > $this->options['timeout']) {
                Log::warning('split timeout');

                return true;
            }
        }

        return false;
    }

    /**
     * 核心拆分函数
     *
     * @param  array  $node  word to be look up 要查询的词
     * @param  int  $deep  当前递归深度
     * @param  bool  $forward  搜索方向 true 正向  false 反向
     * @param  bool  $express=true,  快速查询
     * @param  int  $adj_len=0  长度校正系数
     * @param  int  $c_threshhold  信心指数阈值
     * @return void
     */
    private function split(&$node, $deep = 0, $express = false, $adj_len = 0, $c_threshhold = 0.8, $w_threshhold = 0.8, $forward = true, $sandhi_advance = false)
    {
        $strWord = $node['remain'];
        $this->log("spliting word={$strWord} deep={$deep}");
        $output = [];
        // currPathCf是当前搜索路径信心指数，如果过低，马上终止这个路径的搜索
        if ($deep == 0) {
            $this->currPathCf = 1;
        }
        // 达到最大搜索深度，返回
        if ($deep >= $this->MAX_DEEP) {
            return;
        }
        // 直接找到

        $confidence = $this->isExist($strWord, $adj_len);
        if ($confidence > $c_threshhold) {
            array_push($output, [$strWord, '', $confidence]);
            if (isset($node['sum_cf'])) {
                $parent_sum_cf = $node['sum_cf'];
            } else {
                $parent_sum_cf = 1;
            }
            $sum_cf = $parent_sum_cf * $confidence;
            $node['children'][] = ['word' => $strWord, 'remain' => '', 'cf' => $confidence, 'sum_cf' => $sum_cf];
            $this->log("直接找到{$strWord}-{$confidence}");
        } elseif (mb_strlen($strWord, 'UTF-8') < 6) {
            // 按照语尾查询
            $search = "[{$strWord}]";
            $confidence = $this->isExist($search);
            $this->log("查询:{$search}-信心指数{$confidence}");
            if ($confidence > $c_threshhold) {
                array_push($output, [$search, '', $confidence]);
                if (isset($node['sum_cf'])) {
                    $parent_sum_cf = $node['sum_cf'];
                } else {
                    $parent_sum_cf = 1;
                }
                $sum_cf = $parent_sum_cf * $confidence;
                $node['children'][] = ['word' => $search, 'remain' => '', 'cf' => $confidence, 'sum_cf' => $sum_cf];
                $this->log("直接找到{$strWord}-{$confidence}");
            }
        }

        // 如果开头有双辅音，去掉第一个辅音。因为巴利语中没有以双辅音开头的单词。
        $doubleword = 'kkggccjjṭṭḍḍttddppbb';
        if (mb_strlen($strWord, 'UTF-8') > 2) {
            $left2 = mb_substr($strWord, 0, 2, 'UTF-8');
            if (mb_strpos($doubleword, $left2, 0, 'UTF-8') !== false) {
                $strWord = mb_substr($strWord, 1, null, 'UTF-8');
            }
        }

        $len = mb_strlen($strWord, 'UTF-8');
        if ($len > 2) {
            if ($forward) {
                // 正向切
                $this->log('正向切');
                for ($i = $len; $i > 1; $i--) {
                    if ($this->isTimeOut()) {
                        Log::warning('line '.__LINE__);

                        return;
                    }
                    // 应用连音规则切分单词
                    foreach ($this->sandhi as $key => $row) {
                        if ($sandhi_advance == false && $row['advance'] == true) {
                            // continue;
                        }
                        if (mb_substr($strWord, $i - $row['len'], $row['len'], 'UTF-8') == $row['c']) {
                            $str1 = mb_substr($strWord, 0, $i - $row['len'], 'UTF-8').$row['a'];
                            $str2 = $row['b'].mb_substr($strWord, $i, null, 'UTF-8');
                            $confidence = $this->isExist($str1, $adj_len) * $row['cf'];
                            if ($confidence > $c_threshhold) {
                                // 信心指数大于预设的阈值，插入
                                array_push($output, [$str1, $str2, $confidence, $row['adj_len']]);
                                if (isset($node['sum_cf'])) {
                                    $parent_sum_cf = $node['sum_cf'];
                                } else {
                                    $parent_sum_cf = 1;
                                }
                                $sum_cf = $parent_sum_cf * $confidence;
                                if ($sum_cf > $c_threshhold) {
                                    $node['children'][] = [
                                        'word' => $str1,
                                        'remain' => $str2,
                                        'cf' => $confidence,
                                        'sum_cf' => $sum_cf,
                                        'children' => [],
                                    ];
                                }

                                $this->log("插入结构数组：{$str1} 剩余{$str2} 应用：{$row['a']}-{$row['b']}-{$row['c']}");
                                if ($express) {
                                    break;
                                }
                            }
                        }
                    }
                }
            } else {
                // 反向切
                for ($i = 1; $i < $len - 1; $i++) {
                    foreach ($this->sandhi as $key => $row) {
                        if ($this->isTimeOut()) {
                            Log::warning('line '.__LINE__);

                            return;
                        }
                        if ($sandhi_advance == false && $row['advance'] == true) {
                            // continue;
                        }
                        if (mb_substr($strWord, $i, $row['len'], 'UTF-8') == $row['c']) {
                            $str1 = mb_substr($strWord, 0, $i, 'UTF-8').$row['a'];
                            $str2 = $row['b'].mb_substr($strWord, $i + $row['len'], null, 'UTF-8');
                            $confidence = $this->isExist($str2, $adj_len) * $row['cf'];
                            if ($confidence > $c_threshhold) {
                                array_push($output, [$str2, $str1, $confidence, $row['adj_len']]);
                                if (isset($node['sum_cf'])) {
                                    $parent_sum_cf = $node['sum_cf'];
                                } else {
                                    $parent_sum_cf = 1;
                                }
                                $sum_cf = $parent_sum_cf * $confidence;
                                if ($sum_cf > $c_threshhold) {
                                    $node['children'][] = [
                                        'word' => $str2,
                                        'remain' => $str1,
                                        'cf' => $confidence,
                                        'sum_cf' => $sum_cf,
                                        'children' => [],
                                    ];
                                }
                                $this->log("将此次结果插入结果数组：剩余={$str2}");
                                if ($express) {
                                    break;
                                }
                            }
                        }
                    }
                }
            }
        }

        $word = '';
        $this->log('结果数组个数：'.count($output));
        // print_r($node);
        // 遍历children
        foreach ($node['children'] as $key => $child) {
            if ($this->isTimeOut()) {
                Log::warning('line '.__LINE__);

                return;
            }
            // code...
            if (isset($child) && ! empty($child['remain'])) {
                $this->split($node['children'][$key], ($deep + 1), $express, $adj_len, $c_threshhold, $w_threshhold, $forward, $sandhi_advance);
            }
        }
    }

    /**
     * 遍历树状结构，获取结果
     */
    private function get_result($node, &$path)
    {

        // code...
        $path[] = $node['word'];
        if (isset($node['children']) && count($node['children']) > 0) {
            foreach ($node['children'] as $key => $value) {
                $this->get_result($value, $path);
            }
        } else {
            if (empty($node['remain'])) {
                $factors = trim(implode('+', $path), '+');
                $this->result[$factors] = $node['sum_cf'];
            } else {
            }
        }
        array_pop($path);
    }

    /**
     * 颠倒词序
     */
    public function word_reverse($word)
    {
        $reverse = [];
        $newword = explode('+', $word);
        $len = count($newword);
        if ($len > 0) {
            for ($i = $len - 1; $i >= 0; $i--) {
                // code...
                $reverse[] = $newword[$i];
            }
            $output = implode('+', $reverse);

            return $output;
        } else {
            return $word;
        }
    }

    /**
     * 拆分后的处理
     */
    public function split2($word)
    {
        $input = explode('+', $word);
        $newword = [];
        foreach ($input as $value) {
            // 去掉带小括号的调试信息
            $word = strstr($value, '(', true);
            if ($word == false) {
                $word = $value;
            }
            if (mb_strlen($word, 'UTF-8') > 4) {
                // 先看有没有中文意思
                // $this->log("先看有没有中文意思");
                if (UserDict::where('word', $word)->where('mean', '<>', '')->where('language', '<>', 'my')->exists()) {
                    $newword[] = $word;
                } else {
                    // $this->log("如果没有查巴缅替换拆分");
                    // 如果没有查巴缅替换拆分
                    if (UserDict::where('word', $word)->where('dict_id', '61f23efb-b526-4a8e-999e-076965034e60')->exists()) {
                        $pmPart = explode('+', UserDict::where('word', $word)->where('dict_id', '61f23efb-b526-4a8e-999e-076965034e60')->value('factors'));
                        foreach ($pmPart as $pm) {
                            // code...
                            $newword[] = $pm;
                        }
                    } else {
                        // $this->log("如果没有查规则变形");
                        // 如果没有查规则变形
                        if (UserDict::where('word', $word)->where('source', '_SYS_REGULAR_')->exists()) {
                            $rglPart = explode('+', UserDict::where('word', $word)->where('source', '_SYS_REGULAR_')->value('factors'));
                            // 看巴缅有没有第一部分
                            // $this->log("看巴缅有没有第一部分");
                            if (UserDict::where('word', $rglPart[0])->where('dict_id', '61f23efb-b526-4a8e-999e-076965034e60')->exists()) {
                                $pmPart = explode('+', UserDict::where('word', $rglPart[0])->where('dict_id', '61f23efb-b526-4a8e-999e-076965034e60')->value('factors'));
                                foreach ($pmPart as $pm) {
                                    // code...
                                    $newword[] = $pm;
                                }
                            } else {
                                // 没有
                                $newword[] = $rglPart[0];
                            }
                            $newword[] = $rglPart[1];
                        } else {
                            // 还没有就认命了
                            // $this->log("还没有就认命了");
                            $newword[] = $word;
                        }
                    }
                }
            } else {
                $newword[] = $word;
            }
        }

        return implode('+', $newword);
    }

    /**
     * 预处理连音词
     */
    public function splitSandhi($word)
    {
        $newWord = '';
        $firstWord = $word;
        do {
            $isFound = false;
            foreach ($this->sandhi2 as $key => $sandhi) {
                // code...
                $len = $sandhi['len'];
                $end = mb_substr($firstWord, 0 - $len, null, 'UTF-8');
                if ($end == $sandhi['c']) {
                    $word1 = mb_substr($firstWord, 0, mb_strlen($firstWord, 'UTF-8') - $len, 'UTF-8').$sandhi['a'];
                    $word2 = $sandhi['b'];
                    $newWord = $word2.'-'.$newWord;
                    $firstWord = $word1;
                    $isFound = true;
                    break;
                }
            }
        } while ($isFound);
        $newWord = $firstWord.'-'.$newWord;

        return mb_substr($newWord, 0, -1, 'UTF-8');
    }

    /**
     * 切分函数
     *
     * @param  string  $word  需要切分的单词
     * @return array
     */
    public function splitA($word)
    {
        $this->started_at = time();
        $caseman = new CaseMan;
        $output = [];
        // 预处理连音词
        $word1 = $this->splitSandhi($word);
        // 处理双元音
        $this->log('处理双元音');
        $arrword = $this->splitDiphthong($word1);
        if (count($arrword) > 1) {
            array_push($output, [
                'word' => $word,
                'type' => '.un.',
                'grammar' => '',
                'parent' => '',
                'factors' => implode('+', $arrword),
                'confidence' => 0.9999,
            ]);
        }

        foreach ($arrword as $oneword) {
            if (mb_strlen($oneword) < 5) {
                continue;
            }
            $this->result = []; // 清空递归程序的输出容器
            $node = ['word' => '', 'remain' => $oneword, 'children' => []];
            if (mb_strlen($oneword) > 35) {
                // 长词使用快速切分 正向切分 不使用少见sandi规则
                $this->split($node, 0, true, 0.8, 0.9, 0, true, false);
                $min_result = 1;
            } else {
                $this->split($node, 0, false, 0.8, 0.9, 0, true, false);
                $min_result = 2;
            }
            $path = [];
            $this->log($node);
            $this->get_result($node, $path);

            $this->log('正向切分结束 结果数量'.count($this->result));

            if (count($this->result) < $min_result) {
                // 有效结果过少
                $node = ['word' => '', 'remain' => $oneword, 'children' => []];
                $this->split($node, 0, false, 0.2, 0.8, 0, true, true);
                $this->log('有效结果过少 再次正切'.count($this->result));
                if (count($this->result) < 2) {
                    $node = ['word' => '', 'remain' => $oneword, 'children' => []];
                    $this->split($node, 0, false, 0.2, 0.8, 0, false, true);
                    $this->log('有效结果过少 再次反切：结果数量'.count($this->result));
                }
            }

            $this->log("{$oneword}:".count($this->result));
            if (count($this->result) > 0) {
                arsort($this->result); // 按信心指数升序排序
                $iCount = 0;
                foreach ($this->result as $row => $value) {
                    $factors = $row;
                    if (strpos($row, ']+') !== false) {
                        $type = '.un.';
                        $factors = \str_replace(['+[ṃ]+', '[ṃ]+'], 'ṃ+', $row);
                    } else {
                        $type = '.cp.';
                    }
                    $newword = ['word' => $oneword, 'type' => $type, 'grammar' => '', 'parent' => '', 'factors' => $factors, 'confidence' => $value];
                    array_push($output, $newword);

                    if ($iCount == 0) {
                        // 对于最优结果进行处理 找到base
                        $wordWithType = ['word' => $oneword, 'type' => '', 'grammar' => '', 'parent' => '', 'factors' => $factors, 'confidence' => $value];

                        $this->log('查找base');

                        $factors = explode('+', $row);
                        $endOfFactor = end($factors);
                        if (strpos($endOfFactor, '[') !== false) {
                            if (count($factors) >= 2) {
                                $endOfFactor = $factors[count($factors) - 2];
                            }
                        }
                        $this->log('结尾词：'.$endOfFactor);

                        // 猜测单词的base
                        $parents = $caseman->WordToBase($oneword, 1, false);
                        // 找到结尾单词的base
                        $end_parents = $caseman->WordToBase($endOfFactor);

                        if (count($parents) > 0) {
                            foreach ($parents as $base => $case) {
                                // code...
                                if (count($end_parents) > 0) {
                                    foreach ($end_parents as $base2 => $case2) {
                                        if (\mb_substr($base2, -2) === \mb_substr($base, -2)) {
                                            $this->log("{$base} ok");
                                            foreach ($case as $value) {
                                                // code...
                                                foreach ($case2 as $value2) {
                                                    // 验证语法信息是否正确
                                                    if (
                                                        $value['type'] == $value2['type'] &&
                                                        substr($value['grammar'], 0, 3) === substr($value2['grammar'], 0, 3) &&
                                                        $value['confidence'] > 0.5
                                                    ) {
                                                        $wordWithType['type'] = $value['type'];
                                                        $wordWithType['grammar'] = $value['grammar'];
                                                        $wordWithType['factors'] = $value['factors'];
                                                        $wordWithType['parent'] = $base;
                                                        $wordWithType['confidence'] = $value2['confidence'];
                                                        $this->log("word:{$wordWithType['word']} ; type:{$wordWithType['type']}; grammar:{$wordWithType['grammar']};parent:{$wordWithType['parent']}");
                                                        array_push($output, $wordWithType);
                                                    }
                                                }
                                            }
                                        }
                                    }
                                } else {
                                    foreach ($case as $value) {
                                        $wordWithType['type'] = $value['type'];
                                        $wordWithType['grammar'] = $value['grammar'];
                                        $wordWithType['factors'] = $value['factors'];
                                        $wordWithType['parent'] = $base;
                                        $wordWithType['confidence'] = 0.1;
                                        array_push($output, $wordWithType);
                                    }
                                }
                            }
                        }
                    }
                    // 后处理 进一步切分没有意思的长词
                    $this->log('后处理 进一步切分没有意思的长词');
                    $new = $this->split2($row);
                    if ($new !== $row) {
                        $newword['factors'] = $new;
                        array_push($output, $newword);
                        // 再处理一次
                        $new2 = $this->split2($new);
                        if ($new2 !== $new) {
                            $newword['factors'] = $new2;
                            array_push($output, $newword);
                        }
                    }
                    $iCount++;
                    if ($iCount > $this->MAX_RESULT2) {
                        break;
                    }
                }
            } else {
                $this->log("{$oneword} 切分失败");
                $this->log('猜测可能的格位');
                // 猜测单词的base
                $wordWithType = ['word' => $oneword, 'type' => '', 'grammar' => '', 'parent' => '', 'factors' => '', 'confidence' => 0];
                $parents = $caseman->WordToBase($oneword, 1, false);
                foreach ($parents as $base => $case) {
                    foreach ($case as $value) {
                        $wordWithType['type'] = $value['type'];
                        $wordWithType['grammar'] = $value['grammar'];
                        $wordWithType['factors'] = $value['factors'];
                        $wordWithType['parent'] = $base;
                        $wordWithType['confidence'] = $value['confidence'];
                        $this->log("word:{$wordWithType['word']} ; type:{$wordWithType['type']}; grammar:{$wordWithType['grammar']};parent:{$wordWithType['parent']}");
                        array_push($output, $wordWithType);
                    }
                }
            }
        }

        return $output;
    }

    public function setting($param = null) {}

    public function getResult()
    {
        return $this->result;
    }

    public function debug($debug)
    {
        $this->isDebug = $debug;
    }

    private function log($message)
    {
        if ($this->isDebug) {
            Log::info($message);
        }
    }

    private function pushResult($word, $cf)
    {
        array_push($this->result, [$word => $cf]);
    }
}