turbo_split.php 28 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620
  1. <?php
  2. require_once '../public/casesuf.inc';
  3. //require_once '../studio/dict_find_un.inc';
  4. //require_once '../studio/sandhi.php';
  5. require_once "../path.php";
  6. require_once "../public/_pdo.php";
  7. require_once "../redis/function.php";
  8. global $redis;
  9. $redis = redis_connect();
  10. // open word part db
  11. global $dbh;
  12. $dns = "" . _FILE_DB_PART_;
  13. $dbh = new PDO($dns, "", "", array(PDO::ATTR_PERSISTENT => true));
  14. $dbh->setAttribute(PDO::ATTR_ERRMODE, PDO::ERRMODE_WARNING);
  15. global $path;
  16. global $confidence;
  17. global $result;
  18. global $part;
  19. define("MAX_RESULT",100);
  20. $part = array();
  21. $path[] = array("", 0);
  22. $path[] = array("", 0);
  23. $path[] = array("", 0);
  24. $path[] = array("", 0);
  25. $path[] = array("", 0);
  26. $path[] = array("", 0);
  27. $path[] = array("", 0);
  28. $path[] = array("", 0);
  29. $path[] = array("", 0);
  30. $path[] = array("", 0);
  31. $path[] = array("", 0);
  32. $path[] = array("", 0);
  33. $path[] = array("", 0);
  34. $path[] = array("", 0);
  35. $path[] = array("", 0);
  36. $path[] = array("", 0);
  37. $path[] = array("", 0);
  38. global $sandhi;
  39. //sandhi rules table 语尾表
  40. $sandhi[] = array("a" => "", "b" => "", "c" => "", "len" => 0, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  41. $sandhi[] = array("a" => "a", "b" => "a", "c" => "ā", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  42. $sandhi[] = array("a" => "ā", "b" => "ā", "c" => "ā", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  43. $sandhi[] = array("a" => "a", "b" => "ā", "c" => "ā", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  44. $sandhi[] = array("a" => "ā", "b" => "a", "c" => "ā", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  45. $sandhi[] = array("a" => "a", "b" => "e", "c" => "e", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  46. $sandhi[] = array("a" => "a", "b" => "i", "c" => "i", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  47. $sandhi[] = array("a" => "a", "b" => "o", "c" => "o", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  48. $sandhi[] = array("a" => "a", "b" => "u", "c" => "o", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  49. $sandhi[] = array("a" => "u", "b" => "a", "c" => "o", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  50. $sandhi[] = array("a" => "u", "b" => "u", "c" => "ū", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  51. $sandhi[] = array("a" => "a", "b" => "u", "c" => "u", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  52. $sandhi[] = array("a" => "a", "b" => "ī", "c" => "ī", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  53. $sandhi[] = array("a" => "a", "b" => "ū", "c" => "ū", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  54. $sandhi[] = array("a" => "a", "b" => "i", "c" => "e", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  55. $sandhi[] = array("a" => "e", "b" => "a", "c" => "e", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  56. $sandhi[] = array("a" => "i", "b" => "i", "c" => "ī", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  57. $sandhi[] = array("a" => "i", "b" => "e", "c" => "e", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  58. $sandhi[] = array("a" => "i", "b" => "a", "c" => "ya", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  59. $sandhi[] = array("a" => "a", "b" => "atth", "c" => "atth", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  60. $sandhi[] = array("a" => "taṃ", "b" => "n", "c" => "tann", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  61. $sandhi[] = array("a" => "[ṃ]", "b" => "api", "c" => "mpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  62. $sandhi[] = array("a" => "[ṃ]", "b" => "eva", "c" => "meva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  63. $sandhi[] = array("a" => "[o]", "b" => "iva", "c" => "ova", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  64. $sandhi[] = array("a" => "o", "b" => "a", "c" => "o", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  65. $sandhi[] = array("a" => "a", "b" => "ādi", "c" => "ādi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  66. $sandhi[] = array("a" => "a[ānaṃ]", "b" => "a", "c" => "ānama", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  67. $sandhi[] = array("a" => "a", "b" => "iti", "c" => "āti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  68. $sandhi[] = array("a" => "[ṃ]", "b" => "ca", "c" => "ñca", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  69. $sandhi[] = array("a" => "[ṃ]", "b" => "iti", "c" => "nti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  70. $sandhi[] = array("a" => "[ṃ]", "b" => "a", "c" => "ma", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  71. $sandhi[] = array("a" => "ṃ", "b" => "a", "c" => "m", "len" => 1, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  72. $sandhi[] = array("a" => "[ṃ]", "b" => "ā", "c" => "mā", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  73. $sandhi[] = array("a" => "[ṃ]", "b" => "u", "c" => "mu", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  74. $sandhi[] = array("a" => "[ṃ]", "b" => "h", "c" => "ñh", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  75. $sandhi[] = array("a" => "ā", "b" => "[ṃ]", "c" => "am", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  76. $sandhi[] = array("a" => "a", "b" => "[ṃ]", "c" => "am", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  77. $sandhi[] = array("a" => "ī", "b" => "[ṃ]", "c" => "im", "len" => 2, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  78. $sandhi[] = array("a" => "ati", "b" => "tabba", "c" => "atabba", "len" => 6, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  79. $sandhi[] = array("a" => "ati", "b" => "tabba", "c" => "itabba", "len" => 6, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  80. $sandhi[] = array("a" => "iti", "b" => "a", "c" => "icca", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  81. $sandhi[] = array("a" => "uṃ", "b" => "a", "c" => "uma", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  82. $sandhi[] = array("a" => "u[ūnaṃ]", "b" => "a", "c" => "ūnama", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  83. $sandhi[] = array("a" => "ī[īnaṃ]", "b" => "a", "c" => "īnama", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  84. $sandhi[] = array("a" => "su", "b" => "a", "c" => "sva", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  85. #other sandhi rule. can be use but program will be slow down
  86. #其他连音规则,如果使用则会让程序运行变慢
  87. $sandhi[] = array("a" => "ā", "b" => "iti", "c" => "āti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  88. $sandhi[] = array("a" => "a", "b" => "iti", "c" => "āti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  89. $sandhi[] = array("a" => "e", "b" => "iti", "c" => "eti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  90. $sandhi[] = array("a" => "ī", "b" => "iti", "c" => "īti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  91. $sandhi[] = array("a" => "i", "b" => "iti", "c" => "īti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  92. $sandhi[] = array("a" => "o", "b" => "iti", "c" => "oti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  93. $sandhi[] = array("a" => "ū", "b" => "iti", "c" => "ūti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  94. $sandhi[] = array("a" => "u", "b" => "iti", "c" => "ūti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  95. $sandhi[] = array("a" => "ṃ", "b" => "iti", "c" => "nti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  96. $sandhi[] = array("a" => "ṃ", "b" => "ca", "c" => "ñca", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  97. $sandhi[] = array("a" => "ṃ", "b" => "cāti", "c" => "ñcāti", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  98. $sandhi[] = array("a" => "ṃ", "b" => "cet", "c" => "ñcet", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  99. $sandhi[] = array("a" => "ṃ", "b" => "ev", "c" => "mev", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  100. /*
  101. $sandhi2[] = array("a" => "ṃ", "b" => "ca", "c" => "ñca", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  102. $sandhi2[] = array("a" => "ṃ", "b" => "hi", "c" => "ñhi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  103. $sandhi2[] = array("a" => "a", "b" => "iti", "c" => "āti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  104. $sandhi2[] = array("a" => "e", "b" => "iti", "c" => "eti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  105. $sandhi2[] = array("a" => "i", "b" => "iti", "c" => "īti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  106. $sandhi2[] = array("a" => "o", "b" => "iti", "c" => "oti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  107. $sandhi2[] = array("a" => "u", "b" => "iti", "c" => "ūti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  108. $sandhi2[] = array("a" => "ṃ", "b" => "iti", "c" => "nti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  109. $sandhi2[] = array("a" => "a", "b" => "eva", "c" => "eva", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  110. $sandhi2[] = array("a" => "e", "b" => "eva", "c" => "eva", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  111. $sandhi2[] = array("a" => "i", "b" => "eva", "c" => "yeva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  112. $sandhi2[] = array("a" => "ī", "b" => "eva", "c" => "iyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  113. $sandhi2[] = array("a" => "ī", "b" => "eva", "c" => "īyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  114. $sandhi2[] = array("a" => "o", "b" => "eva", "c" => "ova", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  115. $sandhi2[] = array("a" => "u", "b" => "eva", "c" => "veva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  116. $sandhi2[] = array("a" => "ā", "b" => "api", "c" => "āpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  117. $sandhi2[] = array("a" => "a", "b" => "api", "c" => "āpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  118. $sandhi2[] = array("a" => "e", "b" => "api", "c" => "epi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  119. $sandhi2[] = array("a" => "ī", "b" => "api", "c" => "īpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  120. $sandhi2[] = array("a" => "i", "b" => "api", "c" => "īpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  121. $sandhi2[] = array("a" => "o", "b" => "api", "c" => "opi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  122. $sandhi2[] = array("a" => "ū", "b" => "api", "c" => "ūpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  123. $sandhi2[] = array("a" => "u", "b" => "api", "c" => "ūpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  124. $sandhi2[] = array("a" => "u", "b" => "api", "c" => "upi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  125. $sandhi2[] = array("a" => "ṃ", "b" => "api", "c" => "mpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  126. */
  127. $sandhi[] = array("a" => "a", "b" => "a", "c" => "a", "len" => 1, "adj_len" => -1, "advance" => true,"cf"=>0.99);
  128. $sandhi[] = array("a" => "ī", "b" => "", "c" => "i", "len" => 1, "adj_len" => 0, "advance" => true,"cf"=>0.9);
  129. function split_diphthong($word)
  130. {
  131. //diphthong table双元音表
  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. //将双元音拆开
  135. //step 1 : split at diphthong . ~aa~ -> ~a-a~
  136. $word1 = str_replace($search, $replace, $word);
  137. //按连字符拆开处理
  138. $arrword = str_getcsv($word1, "-");
  139. return $arrword;
  140. }
  141. /*
  142. 用于数组连接字符串
  143. */
  144. function myfunction($v1, $v2)
  145. {
  146. return $v1 . "+" . $v2;
  147. }
  148. function microtime_float()
  149. {
  150. list($usec, $sec) = explode(" ", microtime());
  151. return ((float) $usec + (float) $sec);
  152. }
  153. function dict_lookup($word)
  154. {
  155. if (strlen($word) <= 1) {
  156. return 0;
  157. }
  158. global $case;
  159. global $dbh;
  160. $str = strstr($word, "[");
  161. if ($str === false) {
  162. $search = $word;
  163. } else {
  164. $search = $str;
  165. }
  166. $query = "SELECT weight from part where word = ? ";
  167. $stmt = $dbh->prepare($query);
  168. $stmt->execute(array($search));
  169. $row = $stmt->fetch(PDO::FETCH_NUM);
  170. if ($row) {
  171. return array($row[0],0);
  172. } else {
  173. //去除尾查
  174. $newWord = array();
  175. for ($row = 0; $row < count($case); $row++) {
  176. $len = mb_strlen($case[$row][1], "UTF-8");
  177. $end = mb_substr($word, 0 - $len, null, "UTF-8");
  178. if ($end == $case[$row][1]) {
  179. $base = mb_substr($word, 0, mb_strlen($word, "UTF-8") - $len, "UTF-8") . $case[$row][0];
  180. if ($base != $word) {
  181. $newWord[$base] = mb_strlen($case[$row][1],"UTF-8");
  182. }
  183. }
  184. }
  185. #找到最高频的base
  186. $base_weight = 0;
  187. $len = 0;
  188. foreach ($newWord as $x => $x_value) {
  189. $query = "SELECT weight from part where word = ? ";
  190. $stmt = $dbh->prepare($query);
  191. $stmt->execute(array($x));
  192. $row = $stmt->fetch(PDO::FETCH_NUM);
  193. if ($row) {
  194. if ($row[0] > $base_weight) {
  195. $base_weight = $row[0];
  196. $len=$x_value;
  197. }
  198. }
  199. }
  200. return array($base_weight,$len);
  201. }
  202. }
  203. /*
  204. 查找某个单词是否在现有词典出现
  205. 返回信心指数
  206. look up single word in dictionary vocabulary
  207. return the confidence value
  208. */
  209. function isExsit($word, $adj_len = 0)
  210. {
  211. global $auto_split_times;
  212. global $result;
  213. global $part;
  214. global $confidence;
  215. $auto_split_times++;
  216. if (isset($_POST["debug"])) {
  217. echo "<div>正在查询:{$word}</div>";
  218. }
  219. $isFound = false;
  220. $count = 0;
  221. if (isset($part["{$word}"])) {
  222. $word_count = $part["{$word}"][0];
  223. $case_len = $part["{$word}"][1];
  224. if ($word_count > 0) {
  225. if (isset($_POST["debug"])) {
  226. echo "查到:{$word}:{$word_count}个\n";
  227. }
  228. $isFound = true;
  229. $count = $word_count + 1;
  230. }
  231. } else {
  232. $db = dict_lookup($word);
  233. $word_count = $db[0];
  234. $case_len = $db[1];
  235. //加入查询缓存
  236. $part["{$word}"] = $db;
  237. if ($word_count > 0) {
  238. if (isset($_POST["debug"])) {
  239. echo "查到:{$word}:{$word_count}个\n";
  240. }
  241. $isFound = true;
  242. $count = $word_count + 1;
  243. }
  244. }
  245. //fomular of confidence value 信心值计算公式
  246. if ($isFound) {
  247. if (isset($confidence["{$word}"])) {
  248. $cf = $confidence["{$word}"];
  249. } else {
  250. //$len = mb_strlen($word, "UTF-8") + $adj_len;
  251. $len = mb_strlen($word, "UTF-8") - $case_len;
  252. $len_correct = 1.2;
  253. $count2 = 1.1 + pow($count, 1.18);
  254. $conf_num = pow(1 / $count2, pow(($len - 0.5), $len_correct));
  255. $cf = round(1 / (1 + 640 * $conf_num), 9);
  256. //$cf = round((1-0.02*$case_len) / (1 + 640 * $conf_num), 9);
  257. $confidence["{$word}"] = $cf;
  258. if (isset($_POST["debug"])) {
  259. echo "信心指数:{$word}:{$cf}\n";
  260. }
  261. }
  262. return ($cf);
  263. } else {
  264. return (-1);
  265. }
  266. }
  267. /*
  268. 核心拆分函数
  269. $strWord, word to be look up 要查询的词
  270. $deep, 当前递归深度
  271. $express=true, 快速查询
  272. $adj_len=0 长度校正系数
  273. $c_threshhold 信心指数阈值
  274. */
  275. function mySplit2($strWord, $deep = 0, $express = false, $adj_len = 0, $c_threshhold = 0.8, $w_threshhold = 0.8, $forward = true, $sandhi_advance = false)
  276. {
  277. global $path;
  278. global $result;
  279. global $sandhi;
  280. $output = array();
  281. //达到最大搜索深度,返回
  282. if ($deep >= 16) {
  283. $word = "";
  284. $cf = 1.0;
  285. for ($i = 0; $i < $deep; $i++) {
  286. if (!empty($path[$i][0])) {
  287. $word .= $path[$i][0] . "+";
  288. if (isset($_POST["debug"])) {
  289. $word .= "(" . $path[$i][1] . ")-";
  290. }
  291. $cf = $cf * $path[$i][1];
  292. }
  293. }
  294. $len = pow(mb_strlen($strWord, "UTF-8"), 3);
  295. $cf += (0 - $len) / ($len + 150);
  296. $word .= "{$strWord}";
  297. if ($forward == true) {
  298. $result[$word] = $cf;
  299. return;
  300. } else {
  301. $reverseWord = word_reverse($word);
  302. $result[$reverseWord] = $cf;
  303. return;
  304. }
  305. }
  306. //直接找到
  307. $confidence = isExsit($strWord, $adj_len);
  308. if ($confidence >= 0) {
  309. $output[] = array($strWord, "", $confidence);
  310. }
  311. else {
  312. $confidence = isExsit("[" . $strWord . "]");
  313. if ($confidence >= 0) {
  314. $output[] = array("[" . $strWord . "]", "", $confidence);
  315. }
  316. }
  317. //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。
  318. $doubleword = "kkggccjjṭṭḍḍttddppbb";
  319. if (mb_strlen($strWord, "UTF-8") > 2) {
  320. $left2 = mb_substr($strWord, 0, 2, "UTF-8");
  321. if (mb_strpos($doubleword, $left2, 0, "UTF-8") !== false) {
  322. $strWord = mb_substr($strWord, 1, null, "UTF-8");
  323. }
  324. }
  325. $len = mb_strlen($strWord, "UTF-8");
  326. if ($len > 2) {
  327. if ($forward) {
  328. #正向切
  329. for ($i = $len; $i > 1; $i--) {
  330. foreach ($sandhi as $key => $row) {
  331. if ($sandhi_advance == false && $row["advance"] == true) {
  332. //continue;
  333. }
  334. if (mb_substr($strWord, $i - $row["len"], $row["len"], "UTF-8") == $row["c"]) {
  335. $str1 = mb_substr($strWord, 0, $i - $row["len"], "UTF-8") . $row["a"];
  336. $str2 = $row["b"] . mb_substr($strWord, $i, null, "UTF-8");
  337. $confidence = isExsit($str1, $adj_len)*$row["cf"];
  338. if ($confidence > $c_threshhold) {
  339. $output[] = array($str1, $str2, $confidence, $row["adj_len"]);
  340. if (isset($_POST["debug"])) {
  341. echo "插入:{$str1} 剩余{$str2} 应用:{$row["a"]}-{$row["b"]}-{$row["c"]}\n";
  342. }
  343. if ($express) {
  344. break;
  345. }
  346. }
  347. }
  348. }
  349. }
  350. } else {
  351. #反向切
  352. for ($i = 1; $i < $len - 1; $i++) {
  353. foreach ($sandhi as $key => $row) {
  354. if ($sandhi_advance == false && $row["advance"] == true) {
  355. //continue;
  356. }
  357. if (mb_substr($strWord, $i, $row["len"], "UTF-8") == $row["c"]) {
  358. $str1 = mb_substr($strWord, 0, $i, "UTF-8") . $row["a"];
  359. $str2 = $row["b"] . mb_substr($strWord, $i + $row["len"], null, "UTF-8");
  360. $confidence = isExsit($str2, $adj_len)*$row["cf"];
  361. if ($confidence > $c_threshhold) {
  362. $output[] = array($str2, $str1, $confidence, $row["adj_len"]);
  363. if (isset($_POST["debug"])) {
  364. echo "插入:{$str2}\n";
  365. }
  366. if ($express) {
  367. break;
  368. }
  369. }
  370. }
  371. }
  372. }
  373. }
  374. }
  375. $word = "";
  376. if (count($output) > 0) {
  377. foreach ($output as $part) {
  378. $checked = $part[0];
  379. $remainder = $part[1];
  380. $path[$deep][0] = $checked;
  381. $path[$deep][1] = $part[2];
  382. if (empty($remainder)) {
  383. #全切完了
  384. $word = "";
  385. $cf = 1.0;
  386. for ($i = 0; $i < $deep; $i++) {
  387. $word .= $path[$i][0];
  388. if (isset($_POST["debug"])) {
  389. $word .= "(" . $path[$i][1] . ")";
  390. }
  391. $word .= "+";
  392. $cf = $cf * $path[$i][1];
  393. }
  394. if (isset($_POST["debug"])) {
  395. $word .= $checked . "({$part[2]})";
  396. } else {
  397. $word .= $checked;
  398. }
  399. $cf = $cf * $part[2];
  400. if ($cf > $w_threshhold) {
  401. if ($forward == true) {
  402. $result[$word] = $cf;
  403. return;
  404. } else {
  405. $reverseWord = word_reverse($word);
  406. $result[$reverseWord] = $cf;
  407. return;
  408. }
  409. }
  410. } else {
  411. #接着切
  412. mySplit2($remainder, ($deep + 1), $express, $adj_len, $c_threshhold, $w_threshhold, $forward, $sandhi_advance);
  413. }
  414. }
  415. }
  416. else {
  417. #尾巴查不到了
  418. $word = "";
  419. $cf = 1.0;
  420. for ($i = 0; $i < $deep; $i++) {
  421. $word .= $path[$i][0];
  422. if (isset($_POST["debug"])) {
  423. $word .= "(" . $path[$i][1] . ")";
  424. }
  425. $word .= "+";
  426. $cf = $cf * $path[$i][1];
  427. }
  428. $len = pow(mb_strlen($strWord, "UTF-8"), 3);
  429. if ($forward) {
  430. $cf =(1-$cf) * $len / ($len + 150);
  431. } else {
  432. $cf =(1-$cf) * $len / ($len + 5);
  433. }
  434. if (isset($_POST["debug"])) {
  435. $word = $word.$strWord . "(0)";
  436. } else {
  437. $word = $word .$strWord;
  438. }
  439. if ($cf > $w_threshhold) {
  440. if ($forward == true) {
  441. $result[$word] = $cf;
  442. return;
  443. }
  444. else {
  445. $reverseWord = word_reverse($word);
  446. $result[$reverseWord] = $cf;
  447. return;
  448. }
  449. }
  450. }
  451. }
  452. function word_reverse($word)
  453. {
  454. $reverse = array();
  455. $newword = explode("+", $word);
  456. $len = count($newword);
  457. if ($len > 0) {
  458. for ($i = $len - 1; $i >= 0; $i--) {
  459. # code...
  460. $reverse[] = $newword[$i];
  461. }
  462. $output = implode("+", $reverse);
  463. return $output;
  464. } else {
  465. return $word;
  466. }
  467. }
  468. #后处理
  469. function split2($word){
  470. global $redis;
  471. $input = explode("+",$word);
  472. $newword=array();
  473. foreach ($input as $value) {
  474. $word = strstr($value,"(",true);
  475. if($word==false){
  476. $word=$value;
  477. }
  478. if(mb_strlen($word,"UTF-8")>4){
  479. # 先看有没有中文意思
  480. if($redis->hExists("dict://ref/has_mean",$word)===TRUE){
  481. $newword[]=$word;
  482. }
  483. else{
  484. #如果没有查巴缅替换拆分
  485. if($redis->hExists("dict://pm/part",$word)===TRUE){
  486. $pmPart = explode("+",$redis->hGet("dict://pm/part",$word)) ;
  487. foreach ($pmPart as $pm) {
  488. # code...
  489. $newword[]=$pm;
  490. }
  491. }
  492. else{
  493. #如果没有查规则变形
  494. if($redis->hExists("dict://regular/part",$word)===TRUE){
  495. $rglPart = explode("+",$redis->hGet("dict://regular/part",$word)) ;
  496. #看巴缅有没有第一部分
  497. if($redis->hExists("dict://pm/part",$rglPart[0])===TRUE){
  498. $pmPart = explode("+",$redis->hGet("dict://pm/part",$rglPart[0])) ;
  499. foreach ($pmPart as $pm) {
  500. # code...
  501. $newword[]=$pm;
  502. }
  503. }
  504. else{
  505. #没有
  506. $newword[]=$rglPart[0];
  507. }
  508. $newword[]=$rglPart[1];
  509. }
  510. else{
  511. #还没有就认命了
  512. $newword[]=$word;
  513. }
  514. }
  515. }
  516. }
  517. else{
  518. $newword[]=$word;
  519. }
  520. }
  521. return implode("+",$newword);
  522. }
  523. function preSandhi($word){
  524. $sandhi2[] = array("a" => "ṃ", "b" => "ca", "c" => "ñca", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  525. $sandhi2[] = array("a" => "ṃ", "b" => "hi", "c" => "ñhi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>1.0);
  526. $sandhi2[] = array("a" => "a", "b" => "iti", "c" => "āti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  527. $sandhi2[] = array("a" => "e", "b" => "iti", "c" => "eti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  528. $sandhi2[] = array("a" => "i", "b" => "iti", "c" => "īti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  529. $sandhi2[] = array("a" => "o", "b" => "iti", "c" => "oti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  530. $sandhi2[] = array("a" => "u", "b" => "iti", "c" => "ūti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  531. $sandhi2[] = array("a" => "ṃ", "b" => "iti", "c" => "nti", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.99999);
  532. $sandhi2[] = array("a" => "ī", "b" => "eva", "c" => "iyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  533. $sandhi2[] = array("a" => "ī", "b" => "eva", "c" => "īyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  534. $sandhi2[] = array("a" => "u", "b" => "eva", "c" => "uyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  535. $sandhi2[] = array("a" => "ṃ", "b" => "eva", "c" => "ṃyeva", "len" => 5, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  536. $sandhi2[] = array("a" => "i", "b" => "eva", "c" => "yeva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  537. $sandhi2[] = array("a" => "o", "b" => "eva", "c" => "ova", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  538. $sandhi2[] = array("a" => "ṃ", "b" => "eva", "c" => "meva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  539. $sandhi2[] = array("a" => "u", "b" => "eva", "c" => "veva", "len" => 4, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  540. $sandhi2[] = array("a" => "a", "b" => "eva", "c" => "eva", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  541. $sandhi2[] = array("a" => "e", "b" => "eva", "c" => "eva", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  542. $sandhi2[] = array("a" => "a", "b" => "api", "c" => "āpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  543. $sandhi2[] = array("a" => "ā", "b" => "api", "c" => "āpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  544. $sandhi2[] = array("a" => "e", "b" => "api", "c" => "epi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  545. $sandhi2[] = array("a" => "i", "b" => "api", "c" => "īpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  546. $sandhi2[] = array("a" => "ī", "b" => "api", "c" => "īpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  547. $sandhi2[] = array("a" => "o", "b" => "api", "c" => "opi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  548. $sandhi2[] = array("a" => "u", "b" => "api", "c" => "ūpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  549. $sandhi2[] = array("a" => "ū", "b" => "api", "c" => "ūpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  550. $sandhi2[] = array("a" => "u", "b" => "api", "c" => "upi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  551. $sandhi2[] = array("a" => "ṃ", "b" => "api", "c" => "mpi", "len" => 3, "adj_len" => 0, "advance" => false,"cf"=>0.9999);
  552. $newWord = "";
  553. $firstWord=$word;
  554. do {
  555. $isFound = false;
  556. foreach ($sandhi2 as $key => $sandhi) {
  557. # code...
  558. $len = $sandhi["len"];
  559. $end = mb_substr($firstWord, 0 - $len, null, "UTF-8");
  560. if ($end == $sandhi["c"]) {
  561. $word1 = mb_substr($firstWord, 0, mb_strlen($firstWord, "UTF-8") - $len, "UTF-8") .$sandhi["a"];
  562. $word2 = $sandhi["b"];
  563. $newWord = $word2 . "+" .$newWord;
  564. $firstWord = $word1;
  565. $isFound=true;
  566. break;
  567. }
  568. }
  569. } while ($isFound);
  570. $newWord = $firstWord . "+" .$newWord;
  571. return mb_substr($newWord,0,-1, "UTF-8");
  572. }