split.php 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436
  1. <?php
  2. //强力拆分复合词
  3. require_once '../public/casesuf.inc';
  4. require_once '../studio/dict_find_un.inc';
  5. require_once '../studio/sandhi.php';
  6. require_once "../path.php";
  7. require_once "../public/_pdo.php";
  8. if(isset($_POST["word"])){
  9. $input_word=mb_strtolower(trim($_POST["word"]),'UTF-8');
  10. if(trim($input_word)==""){
  11. echo "Empty";
  12. return;
  13. }
  14. $arrWords = str_getcsv($input_word,"\n");//支持批量拆分
  15. }
  16. else{
  17. ?>
  18. <!--debug only-->
  19. <form action="split.php" method="post">
  20. Words: <textarea type="text" name="word"></textarea>
  21. <input name="debug" />
  22. <input type="submit">
  23. </form>
  24. <?php
  25. return;
  26. }
  27. global $dbh;
  28. $dns = "sqlite:"._FILE_DB_PART_;
  29. $dbh = new PDO($dns, "", "",array(PDO::ATTR_PERSISTENT=>true));
  30. $dbh->setAttribute(PDO::ATTR_ERRMODE, PDO::ERRMODE_WARNING);
  31. global $path;
  32. global $confidence;
  33. global $result;
  34. global $part ;
  35. $part= array();
  36. $path[]=array("",0);
  37. $path[]=array("",0);
  38. $path[]=array("",0);
  39. $path[]=array("",0);
  40. $path[]=array("",0);
  41. $path[]=array("",0);
  42. $path[]=array("",0);
  43. $path[]=array("",0);
  44. $path[]=array("",0);
  45. $path[]=array("",0);
  46. $path[]=array("",0);
  47. $path[]=array("",0);
  48. $path[]=array("",0);
  49. $path[]=array("",0);
  50. $path[]=array("",0);
  51. $path[]=array("",0);
  52. $path[]=array("",0);
  53. global $sandhi ;
  54. //sandhi 语尾表
  55. $sandhi[]=array("a"=>"","b"=>"","c"=>"","len"=>0,"adj_len"=>0,"advance"=>false);
  56. $sandhi[]=array("a"=>"a","b"=>"a","c"=>"ā","len"=>1,"adj_len"=>0,"advance"=>false);
  57. $sandhi[]=array("a"=>"ā","b"=>"ā","c"=>"ā","len"=>1,"adj_len"=>0,"advance"=>false);
  58. $sandhi[]=array("a"=>"a","b"=>"ā","c"=>"ā","len"=>1,"adj_len"=>0,"advance"=>false);
  59. $sandhi[]=array("a"=>"a","b"=>"e","c"=>"e","len"=>1,"adj_len"=>0,"advance"=>false);
  60. $sandhi[]=array("a"=>"a","b"=>"i","c"=>"i","len"=>1,"adj_len"=>0,"advance"=>false);
  61. $sandhi[]=array("a"=>"a","b"=>"o","c"=>"o","len"=>1,"adj_len"=>0,"advance"=>false);
  62. $sandhi[]=array("a"=>"a","b"=>"u","c"=>"o","len"=>1,"adj_len"=>0,"advance"=>false);
  63. $sandhi[]=array("a"=>"u","b"=>"a","c"=>"o","len"=>1,"adj_len"=>0,"advance"=>false);
  64. $sandhi[]=array("a"=>"u","b"=>"u","c"=>"ū","len"=>1,"adj_len"=>0,"advance"=>false);
  65. $sandhi[]=array("a"=>"a","b"=>"u","c"=>"u","len"=>1,"adj_len"=>0,"advance"=>false);
  66. $sandhi[]=array("a"=>"a","b"=>"ī","c"=>"ī","len"=>1,"adj_len"=>0,"advance"=>false);
  67. $sandhi[]=array("a"=>"a","b"=>"ū","c"=>"ū","len"=>1,"adj_len"=>0,"advance"=>false);
  68. $sandhi[]=array("a"=>"a","b"=>"i","c"=>"e","len"=>1,"adj_len"=>0,"advance"=>false);
  69. $sandhi[]=array("a"=>"i","b"=>"i","c"=>"ī","len"=>1,"adj_len"=>0,"advance"=>false);
  70. $sandhi[]=array("a"=>"i","b"=>"e","c"=>"e","len"=>1,"adj_len"=>0,"advance"=>false);
  71. $sandhi[]=array("a"=>"i","b"=>"a","c"=>"ya","len"=>2,"adj_len"=>0,"advance"=>false);
  72. $sandhi[]=array("a"=>"a","b"=>"atth","c"=>"atth","len"=>4,"adj_len"=>0,"advance"=>false);
  73. $sandhi[]=array("a"=>"taṃ","b"=>"n","c"=>"tann","len"=>4,"adj_len"=>0,"advance"=>false);
  74. $sandhi[]=array("a"=>"[ṃ]","b"=>"api","c"=>"mpi","len"=>3,"adj_len"=>0,"advance"=>false);
  75. $sandhi[]=array("a"=>"[ṃ]","b"=>"eva","c"=>"meva","len"=>4,"adj_len"=>0,"advance"=>false);
  76. $sandhi[]=array("a"=>"[o]","b"=>"iva","c"=>"ova","len"=>3,"adj_len"=>0,"advance"=>false);
  77. $sandhi[]=array("a"=>"a","b"=>"ādi","c"=>"ādi","len"=>3,"adj_len"=>0,"advance"=>false);
  78. $sandhi[]=array("a"=>"a[ānaṃ]","b"=>"a","c"=>"ānama","len"=>5,"adj_len"=>0,"advance"=>false);
  79. $sandhi[]=array("a"=>"a","b"=>"iti","c"=>"āti","len"=>3,"adj_len"=>0,"advance"=>false);
  80. $sandhi[]=array("a"=>"[ṃ]","b"=>"ca","c"=>"ñca","len"=>3,"adj_len"=>0,"advance"=>false);
  81. $sandhi[]=array("a"=>"[ṃ]","b"=>"iti","c"=>"nti","len"=>3,"adj_len"=>0,"advance"=>false);
  82. $sandhi[]=array("a"=>"[ṃ]","b"=>"a","c"=>"ma","len"=>2,"adj_len"=>0,"advance"=>false);
  83. $sandhi[]=array("a"=>"[ṃ]","b"=>"ā","c"=>"mā","len"=>2,"adj_len"=>0,"advance"=>false);
  84. $sandhi[]=array("a"=>"[ṃ]","b"=>"u","c"=>"mu","len"=>2,"adj_len"=>0,"advance"=>false);
  85. $sandhi[]=array("a"=>"[ṃ]","b"=>"h","c"=>"ñh","len"=>2,"adj_len"=>0,"advance"=>false);
  86. $sandhi[]=array("a"=>"ā","b"=>"[ṃ]","c"=>"am","len"=>2,"adj_len"=>0,"advance"=>false);
  87. $sandhi[]=array("a"=>"ī","b"=>"[ṃ]","c"=>"im","len"=>2,"adj_len"=>0,"advance"=>false);
  88. $sandhi[]=array("a"=>"ati","b"=>"tabba","c"=>"atabba","len"=>6,"adj_len"=>0,"advance"=>false);
  89. $sandhi[]=array("a"=>"ati","b"=>"tabba","c"=>"itabba","len"=>6,"adj_len"=>0,"advance"=>false);
  90. $sandhi[]=array("a"=>"iti","b"=>"a","c"=>"icca","len"=>4,"adj_len"=>0,"advance"=>false);
  91. /*
  92. $sandhi[]=array("a"=>"u[ūnaṃ]","b"=>"a","c"=>"ūnama","len"=>5,"adj_len"=>0,"advance"=>false);
  93. $sandhi[]=array("a"=>"ī[īnaṃ]","b"=>"a","c"=>"īnama","len"=>5,"adj_len"=>0,"advance"=>false);
  94. $sandhi[]=array("a"=>"ā","b"=>"iti","c"=>"āti","len"=>3,"adj_len"=>0);
  95. $sandhi[]=array("a"=>"a","b"=>"iti","c"=>"āti","len"=>3,"adj_len"=>0);
  96. $sandhi[]=array("a"=>"e","b"=>"iti","c"=>"eti","len"=>3,"adj_len"=>0);
  97. $sandhi[]=array("a"=>"ī","b"=>"iti","c"=>"īti","len"=>3,"adj_len"=>0);
  98. $sandhi[]=array("a"=>"i","b"=>"iti","c"=>"īti","len"=>3,"adj_len"=>0);
  99. $sandhi[]=array("a"=>"o","b"=>"iti","c"=>"oti","len"=>3,"adj_len"=>0);
  100. $sandhi[]=array("a"=>"ū","b"=>"iti","c"=>"ūti","len"=>3,"adj_len"=>0);
  101. $sandhi[]=array("a"=>"u","b"=>"iti","c"=>"ūti","len"=>3,"adj_len"=>0);
  102. $sandhi[]=array("a"=>"ṃ","b"=>"iti","c"=>"nti","len"=>3,"adj_len"=>0);
  103. $sandhi[]=array("a"=>"ṃ","b"=>"ca","c"=>"ñca","len"=>3,"adj_len"=>0);
  104. $sandhi[]=array("a"=>"a","b"=>"eva","c"=>"eva","len"=>3,"adj_len"=>0);
  105. $sandhi[]=array("a"=>"ā","b"=>"eva","c"=>"āyeva","len"=>5,"adj_len"=>0);
  106. $sandhi[]=array("a"=>"e","b"=>"eva","c"=>"eva","len"=>3,"adj_len"=>0);
  107. $sandhi[]=array("a"=>"i","b"=>"eva","c"=>"yeva","len"=>4,"adj_len"=>0);
  108. $sandhi[]=array("a"=>"ī","b"=>"eva","c"=>"iyeva","len"=>5,"adj_len"=>0);
  109. $sandhi[]=array("a"=>"ī","b"=>"eva","c"=>"īyeva","len"=>5,"adj_len"=>0);
  110. $sandhi[]=array("a"=>"o","b"=>"eva","c"=>"ova","len"=>3,"adj_len"=>0);
  111. $sandhi[]=array("a"=>"u","b"=>"eva","c"=>"veva","len"=>3,"adj_len"=>0);
  112. $sandhi[]=array("a"=>"a","b"=>"eva","c"=>"evā","len"=>3,"adj_len"=>0);
  113. $sandhi[]=array("a"=>"e","b"=>"eva","c"=>"evā","len"=>3,"adj_len"=>0);
  114. $sandhi[]=array("a"=>"i","b"=>"eva","c"=>"yevā","len"=>4,"adj_len"=>0);
  115. $sandhi[]=array("a"=>"ī","b"=>"eva","c"=>"yevā","len"=>4,"adj_len"=>0);
  116. $sandhi[]=array("a"=>"ī","b"=>"eva","c"=>"iyevā","len"=>4,"adj_len"=>0);
  117. $sandhi[]=array("a"=>"ī","b"=>"eva","c"=>"īyevā","len"=>4,"adj_len"=>0);
  118. $sandhi[]=array("a"=>"o","b"=>"eva","c"=>"ovā","len"=>4,"adj_len"=>0);
  119. $sandhi[]=array("a"=>"ā","b"=>"api","c"=>"āpi","len"=>3,"adj_len"=>0);
  120. $sandhi[]=array("a"=>"a","b"=>"api","c"=>"āpi","len"=>3,"adj_len"=>0);
  121. $sandhi[]=array("a"=>"e","b"=>"api","c"=>"epi","len"=>3,"adj_len"=>0);
  122. $sandhi[]=array("a"=>"ī","b"=>"api","c"=>"īpi","len"=>3,"adj_len"=>0);
  123. $sandhi[]=array("a"=>"i","b"=>"api","c"=>"īpi","len"=>3,"adj_len"=>0);
  124. $sandhi[]=array("a"=>"o","b"=>"api","c"=>"opi","len"=>3,"adj_len"=>0);
  125. $sandhi[]=array("a"=>"ū","b"=>"api","c"=>"ūpi","len"=>3,"adj_len"=>0);
  126. $sandhi[]=array("a"=>"u","b"=>"api","c"=>"ūpi","len"=>3,"adj_len"=>0);
  127. $sandhi[]=array("a"=>"u","b"=>"api","c"=>"upi","len"=>3,"adj_len"=>0);
  128. $sandhi[]=array("a"=>"ṃ","b"=>"api","c"=>"mpi","len"=>3,"adj_len"=>0);
  129. */
  130. //$sandhi[]=array("a"=>"a","b"=>"a","c"=>"a","len"=>1,"adj_len"=>-1,"advance"=>true);
  131. //$sandhi[]=array("a"=>"ī","b"=>"","c"=>"i","len"=>1,"adj_len"=>0,"advance"=>true);
  132. $search = array('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');
  133. $replace = array('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');
  134. $allword = array();
  135. foreach($arrWords as $oneword){
  136. //预处理
  137. //将双元音拆开
  138. $word = str_replace($search, $replace, $oneword);
  139. if(isset($_POST["debug"])){
  140. echo "Look up:{$word}<br>";
  141. }
  142. //按连字符拆开处理
  143. $arrword = str_getcsv($word,"-");
  144. $t1=microtime_float();
  145. $output = array();
  146. foreach($arrword as $oneword){
  147. $result = array();//全局变量,递归程序的输出容器
  148. if(mb_strlen($oneword,"UTF-8")<30){
  149. mySplit2($oneword,0,true);
  150. }
  151. else{
  152. mySplit2($oneword,0,true);
  153. }
  154. arsort($result);//按信心指数排序
  155. $wordlist = array();
  156. foreach($result as $row=>$value){
  157. $word_part = array();
  158. $word_part["word"] = $row;
  159. $word_part["confidence"] = $value;
  160. $wordlist[] = $word_part;
  161. }
  162. $output[] = $wordlist;
  163. if(isset($_POST["debug"])){
  164. echo "<h2>{$oneword}</h2>";
  165. echo "<h4>".count($result)."</h4>";
  166. }
  167. $iCount=0;
  168. foreach($result as $row=>$value){
  169. if($iCount>10){
  170. break;
  171. }
  172. $iCount++;
  173. $level=$value*90;
  174. if(isset($_POST["debug"])){
  175. echo $row."-[".$value."]<br>";
  176. }
  177. else{
  178. //echo $row."-[".round($level)."] ";
  179. }
  180. }
  181. /*
  182. 后处理
  183. -ssāpi=-[ssa]-api
  184. */
  185. }
  186. $t2 = microtime_float();
  187. $one_split["data"]=$output;
  188. $one_split["time"]= $auto_split_times;
  189. $one_split["second"]= $t2-$t1;
  190. $allword[] = $one_split;
  191. if(isset($_POST["debug"])){
  192. echo "<div>";
  193. echo "<br>查询【{$auto_split_times}】次";
  194. echo "time:".($t2-$t1);
  195. echo "</div>";
  196. }
  197. }
  198. if(isset($_POST["debug"])){
  199. echo "<pre style='margin:2em;padding:1em;background-color:#e9e9e9;'>";
  200. print_r($allword);
  201. echo "</pre>";
  202. }
  203. echo json_encode($allword,JSON_UNESCAPED_UNICODE);
  204. /*
  205. 用于数组连接字符串
  206. */
  207. function myfunction($v1,$v2)
  208. {
  209. return $v1 . "+" . $v2;
  210. }
  211. function microtime_float()
  212. {
  213. list($usec, $sec) = explode(" ", microtime());
  214. return ((float)$usec + (float)$sec);
  215. }
  216. function dict_lookup($word){
  217. global $dbh;
  218. $query = "select weight from part where \"word\" = ? ";
  219. $stmt = $dbh->prepare($query);
  220. $stmt->execute(array($word));
  221. $row = $stmt->fetch(PDO::FETCH_NUM);
  222. if ($row) {
  223. return $row[0];
  224. } else {
  225. return 0;
  226. }
  227. }
  228. /*
  229. 查找某个单词是否在现有词典出现
  230. 返回信心指数
  231. */
  232. function isExsit($word,$adj_len=0){
  233. global $auto_split_times;
  234. global $result;
  235. global $part;
  236. global $confidence;
  237. $auto_split_times++;
  238. //echo "<div>正在查询:{$word}</div>";
  239. $isFound=false;
  240. if(isset($part["{$word}"]))
  241. {
  242. if($part["{$word}"]>0){
  243. $isFound=true;
  244. $count=$part["{$word}"]+1;
  245. }
  246. }
  247. else{
  248. $db=dict_lookup($word);
  249. //加入查询缓存
  250. $part["{$word}"] = $db;
  251. if($db>0){
  252. $isFound=true;
  253. $count=$db+1;
  254. }
  255. }
  256. if($isFound)
  257. {
  258. if(isset($confidence["{$word}"])){
  259. $cf=$confidence["{$word}"];
  260. }
  261. else{
  262. $len=mb_strlen($word,"UTF-8")+$adj_len;
  263. $len_correct=1.2;
  264. $count2=1.1+pow($count,1.18);
  265. $conf_num=pow(1/$count2,pow(($len-1),$len_correct));
  266. $cf=round(1/(1+640*$conf_num),9);
  267. $confidence["{$word}"]=$cf;
  268. }
  269. return($cf);
  270. }
  271. else{
  272. return(-1);
  273. }
  274. }
  275. /*
  276. 核心拆分函数
  277. $strWord, 要查询的词
  278. $deep, 当前递归深度
  279. $turbo=false, 简洁查询
  280. $adj_len=0 长度校正系数
  281. */
  282. function mySplit2($strWord,$deep,$express=false,$adj_len=0,$c_threshhold=0.8){
  283. global $path;
  284. global $result;
  285. global $sandhi ;
  286. $output = array();
  287. //达到最大搜索深度,返回
  288. if($deep>=16){
  289. $word = "";
  290. $cf=1.0;
  291. for($i=0;$i<$deep;$i++){
  292. $word .= $path[$i][0];
  293. if(isset($_POST["debug"])){
  294. $word .="(".$path[$i][1].")-";
  295. }
  296. else{
  297. $word .= "-";
  298. }
  299. $cf=$cf*$path[$i][1];
  300. }
  301. $len=pow(mb_strlen($strWord,"UTF-8"),3);
  302. $cf+=(0-$len)/($len+150);
  303. $word .= "{$strWord}";
  304. $result[$word]=$cf;
  305. return;
  306. }
  307. //直接找到
  308. $confidence=isExsit($strWord,$adj_len);
  309. if($confidence>=0){
  310. $output[] = array($strWord,"",$confidence);
  311. }
  312. else{
  313. $confidence=isExsit("[".$strWord."]");
  314. if($confidence>=0){
  315. $output[] = array("[".$strWord."]","",$confidence);
  316. }
  317. }
  318. //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。
  319. $doubleword="kkggccjjṭṭḍḍttddppbb";
  320. if(mb_strlen($strWord,"UTF-8")>2){
  321. $left2=mb_substr($strWord,0,2,"UTF-8");
  322. if(mb_strpos($doubleword,$left2,0,"UTF-8")!==FALSE){
  323. $strWord=mb_substr($strWord,1,NULL,"UTF-8");
  324. }
  325. }
  326. $len=mb_strlen($strWord,"UTF-8");
  327. if($len>2){
  328. for($i=$len;$i>1;$i--){
  329. foreach($sandhi as $row){
  330. if(mb_substr($strWord,$i-$row["len"],$row["len"],"UTF-8")==$row["c"]){
  331. $str1=mb_substr($strWord,0,$i-$row["len"],"UTF-8").$row["a"];
  332. $str2=$row["b"].mb_substr($strWord,$i,NULL,"UTF-8");
  333. $confidence=isExsit($str1,$adj_len);
  334. if($confidence >= $c_threshhold){
  335. $output[] = array($str1,$str2,$confidence,$row["adj_len"]);
  336. if($express){
  337. break;
  338. }
  339. }
  340. }
  341. }
  342. }
  343. }
  344. if(count($output)>0){
  345. foreach($output as $part){
  346. $path[$deep][0]=$part[0];
  347. $path[$deep][1]=$part[2];
  348. if($part[1]!=""){
  349. mySplit2($part[1],($deep+1),$express,$part[3],$c_threshhold);
  350. }
  351. else{
  352. $word = "";
  353. $cf=1.0;
  354. for($i=0;$i<$deep;$i++){
  355. $word .= $path[$i][0]."+";
  356. if(isset($_POST["debug"])){
  357. $word .= "(".$path[$i][1].")-";
  358. }
  359. $cf=$cf*$path[$i][1];
  360. }
  361. $word .= $part[0];
  362. if(isset($_POST["debug"])){
  363. $word .= "({$part[2]})";
  364. }
  365. $cf=$cf+$part[2]*0.1;
  366. $result[$word]=$cf;
  367. }
  368. }
  369. }
  370. else{
  371. $word = "";
  372. $cf=1.0;
  373. for($i=0;$i<$deep;$i++){
  374. $word .= $path[$i][0]."+";
  375. if(isset($_POST["debug"])){
  376. $word .= "(".$path[$i][1].")-";
  377. }
  378. $cf=$cf*$path[$i][1];
  379. }
  380. $len=pow(mb_strlen($strWord,"UTF-8"),3);
  381. $cf+=(0-$len)/($len+150);
  382. if(isset($_POST["debug"])){
  383. $word .= $strWord."(0)";
  384. }
  385. else{
  386. $word .= $strWord;
  387. }
  388. $result[$word]=$cf;
  389. }
  390. }
  391. ?>