lib/tokeniserlib.php

   1 <?php
   2
   3 //
   4 // NOTICE OF COPYRIGHT
   5 //
   6 // Original code:
   7 //
   8 // Drupal - The copyright of both the Drupal software and the
   9 //          "Druplicon" logo belongs to all the original authors,
  10 //          though both are licensed under the GPL.
  11 //          http://drupal.org
  12 //
  13 // Modifications:
  14 //
  15 // Moodle - Modular Object-Oriented Dynamic Learning Environment
  16 //          http://moodle.com
  17 //
  18 // This file is part of Moodle - http://moodle.org/
  19 //
  20 // Moodle is free software: you can redistribute it and/or modify
  21 // it under the terms of the GNU General Public License as published by
  22 // the Free Software Foundation, either version 3 of the License, or
  23 // (at your option) any later version.
  24 //
  25 // Moodle is distributed in the hope that it will be useful,
  26 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  27 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  28 // GNU General Public License for more details.
  29 //
  30 // You should have received a copy of the GNU General Public License
  31 // along with Moodle.  If not, see <http://www.gnu.org/licenses/>.
  32 //
  33
  34 /**
  35  * Based on Drupal's search.module version 1.224
  36  *
  37  * {@link http://cvs.drupal.org/viewcvs/drupal/drupal/modules/search/search.module?view=markup}
  38  *
  39  * Usage: $tokens = tokenise_text($text)
  40  * Returns an array of tokens (key) with their score (value)
  41  * (see function definition for more info)
  42  *
  43  * Major Contributors
  44  *  - Martin Dougiamas  {@link http://moodle.com}
  45  *  - Eloy Lafuente (stronk7) {@link http://contiento.com}
  46  *  - Antonio Vicent          {@link http://ludens.es}
  47  *
  48  * @package   moodlecore
  49  * @copyright (C) 2001-3001 Eloy Lafuente (stronk7) {@link http://contiento.com}
  50  * @license   http://www.gnu.org/copyleft/gpl.html GNU GPL v3 or later
  51  * @deprecated Moodle 4.5
  52  * @todo MDL-82194 Remove this file.
  53  */
  54
  55 defined('MOODLE_INTERNAL') || die();
  56
  57 debugging(
  58     'The tokeniserlib.php file is deprecated and should not be used any more. ',
  59     DEBUG_DEVELOPER,
  60 );
  61
  62 /**
  63  * Some constants
  64  */
  65 define ('MINIMUM_WORD_SIZE',  3); /// Minimum word size to index and search
  66 define ('MAXIMUM_WORD_SIZE', 50); /// Maximum word size to index and search
  67
  68 define ('START_DELIM',  "\xce\xa9\xc3\x91\xc3\x91"); // Use these (omega) and (ntilde) utf-8 combinations
  69 define ('CENTER_DELIM', "\xc3\x91\xce\xa9\xc3\x91"); // as safe delimiters (not puncuation nor usual) while
  70 define ('END_DELIM',    "\xc3\x91\xc3\x91\xce\xa9"); // processing numbers ($join_numbers = false)
  71
  72 /**
  73  * Matches Unicode character classes to exclude from the search index.
  74  *
  75  * See: http://www.unicode.org/Public/UNIDATA/UCD.html#General_Category_Values
  76  *
  77  * The index only contains the following character classes:
  78  * Lu         Letter, Uppercase
  79  * Ll         Letter, Lowercase
  80  * Lt         Letter, Titlecase
  81  * Lo         Letter, Other
  82  * Nd         Number, Decimal Digit
  83  * No         Number, Other
  84  */
  85 define('PREG_CLASS_SEARCH_EXCLUDE',
  86 '\x{0}-\x{2f}\x{3a}-\x{40}\x{5b}-\x{60}\x{7b}-\x{bf}\x{d7}\x{f7}\x{2b0}-'.
  87 '\x{385}\x{387}\x{3f6}\x{482}-\x{489}\x{559}-\x{55f}\x{589}-\x{5c7}\x{5f3}-'.
  88 '\x{61f}\x{640}\x{64b}-\x{65e}\x{66a}-\x{66d}\x{670}\x{6d4}\x{6d6}-\x{6ed}'.
  89 '\x{6fd}\x{6fe}\x{700}-\x{70f}\x{711}\x{730}-\x{74a}\x{7a6}-\x{7b0}\x{901}-'.
  90 '\x{903}\x{93c}\x{93e}-\x{94d}\x{951}-\x{954}\x{962}-\x{965}\x{970}\x{981}-'.
  91 '\x{983}\x{9bc}\x{9be}-\x{9cd}\x{9d7}\x{9e2}\x{9e3}\x{9f2}-\x{a03}\x{a3c}-'.
  92 '\x{a4d}\x{a70}\x{a71}\x{a81}-\x{a83}\x{abc}\x{abe}-\x{acd}\x{ae2}\x{ae3}'.
  93 '\x{af1}-\x{b03}\x{b3c}\x{b3e}-\x{b57}\x{b70}\x{b82}\x{bbe}-\x{bd7}\x{bf0}-'.
  94 '\x{c03}\x{c3e}-\x{c56}\x{c82}\x{c83}\x{cbc}\x{cbe}-\x{cd6}\x{d02}\x{d03}'.
  95 '\x{d3e}-\x{d57}\x{d82}\x{d83}\x{dca}-\x{df4}\x{e31}\x{e34}-\x{e3f}\x{e46}-'.
  96 '\x{e4f}\x{e5a}\x{e5b}\x{eb1}\x{eb4}-\x{ebc}\x{ec6}-\x{ecd}\x{f01}-\x{f1f}'.
  97 '\x{f2a}-\x{f3f}\x{f71}-\x{f87}\x{f90}-\x{fd1}\x{102c}-\x{1039}\x{104a}-'.
  98 '\x{104f}\x{1056}-\x{1059}\x{10fb}\x{10fc}\x{135f}-\x{137c}\x{1390}-\x{1399}'.
  99 '\x{166d}\x{166e}\x{1680}\x{169b}\x{169c}\x{16eb}-\x{16f0}\x{1712}-\x{1714}'.
 100 '\x{1732}-\x{1736}\x{1752}\x{1753}\x{1772}\x{1773}\x{17b4}-\x{17db}\x{17dd}'.
 101 '\x{17f0}-\x{180e}\x{1843}\x{18a9}\x{1920}-\x{1945}\x{19b0}-\x{19c0}\x{19c8}'.
 102 '\x{19c9}\x{19de}-\x{19ff}\x{1a17}-\x{1a1f}\x{1d2c}-\x{1d61}\x{1d78}\x{1d9b}-'.
 103 '\x{1dc3}\x{1fbd}\x{1fbf}-\x{1fc1}\x{1fcd}-\x{1fcf}\x{1fdd}-\x{1fdf}\x{1fed}-'.
 104 '\x{1fef}\x{1ffd}-\x{2070}\x{2074}-\x{207e}\x{2080}-\x{2101}\x{2103}-\x{2106}'.
 105 '\x{2108}\x{2109}\x{2114}\x{2116}-\x{2118}\x{211e}-\x{2123}\x{2125}\x{2127}'.
 106 '\x{2129}\x{212e}\x{2132}\x{213a}\x{213b}\x{2140}-\x{2144}\x{214a}-\x{2b13}'.
 107 '\x{2ce5}-\x{2cff}\x{2d6f}\x{2e00}-\x{3005}\x{3007}-\x{303b}\x{303d}-\x{303f}'.
 108 '\x{3099}-\x{309e}\x{30a0}\x{30fb}-\x{30fe}\x{3190}-\x{319f}\x{31c0}-\x{31cf}'.
 109 '\x{3200}-\x{33ff}\x{4dc0}-\x{4dff}\x{a015}\x{a490}-\x{a716}\x{a802}\x{a806}'.
 110 '\x{a80b}\x{a823}-\x{a82b}\x{d800}-\x{f8ff}\x{fb1e}\x{fb29}\x{fd3e}\x{fd3f}'.
 111 '\x{fdfc}-\x{fe6b}\x{feff}-\x{ff0f}\x{ff1a}-\x{ff20}\x{ff3b}-\x{ff40}\x{ff5b}-'.
 112 '\x{ff65}\x{ff70}\x{ff9e}\x{ff9f}\x{ffe0}-\x{fffd}');
 113
 114 /**
 115  * Matches all 'N' Unicode character classes (numbers)
 116  */
 117 define('PREG_CLASS_NUMBERS',
 118 '\x{30}-\x{39}\x{b2}\x{b3}\x{b9}\x{bc}-\x{be}\x{660}-\x{669}\x{6f0}-\x{6f9}'.
 119 '\x{966}-\x{96f}\x{9e6}-\x{9ef}\x{9f4}-\x{9f9}\x{a66}-\x{a6f}\x{ae6}-\x{aef}'.
 120 '\x{b66}-\x{b6f}\x{be7}-\x{bf2}\x{c66}-\x{c6f}\x{ce6}-\x{cef}\x{d66}-\x{d6f}'.
 121 '\x{e50}-\x{e59}\x{ed0}-\x{ed9}\x{f20}-\x{f33}\x{1040}-\x{1049}\x{1369}-'.
 122 '\x{137c}\x{16ee}-\x{16f0}\x{17e0}-\x{17e9}\x{17f0}-\x{17f9}\x{1810}-\x{1819}'.
 123 '\x{1946}-\x{194f}\x{2070}\x{2074}-\x{2079}\x{2080}-\x{2089}\x{2153}-\x{2183}'.
 124 '\x{2460}-\x{249b}\x{24ea}-\x{24ff}\x{2776}-\x{2793}\x{3007}\x{3021}-\x{3029}'.
 125 '\x{3038}-\x{303a}\x{3192}-\x{3195}\x{3220}-\x{3229}\x{3251}-\x{325f}\x{3280}-'.
 126 '\x{3289}\x{32b1}-\x{32bf}\x{ff10}-\x{ff19}');
 127
 128 /**
 129  * Matches all 'P' Unicode character classes (punctuation)
 130  */
 131 define('PREG_CLASS_PUNCTUATION',
 132 '\x{21}-\x{23}\x{25}-\x{2a}\x{2c}-\x{2f}\x{3a}\x{3b}\x{3f}\x{40}\x{5b}-\x{5d}'.
 133 '\x{5f}\x{7b}\x{7d}\x{a1}\x{ab}\x{b7}\x{bb}\x{bf}\x{37e}\x{387}\x{55a}-\x{55f}'.
 134 '\x{589}\x{58a}\x{5be}\x{5c0}\x{5c3}\x{5f3}\x{5f4}\x{60c}\x{60d}\x{61b}\x{61f}'.
 135 '\x{66a}-\x{66d}\x{6d4}\x{700}-\x{70d}\x{964}\x{965}\x{970}\x{df4}\x{e4f}'.
 136 '\x{e5a}\x{e5b}\x{f04}-\x{f12}\x{f3a}-\x{f3d}\x{f85}\x{104a}-\x{104f}\x{10fb}'.
 137 '\x{1361}-\x{1368}\x{166d}\x{166e}\x{169b}\x{169c}\x{16eb}-\x{16ed}\x{1735}'.
 138 '\x{1736}\x{17d4}-\x{17d6}\x{17d8}-\x{17da}\x{1800}-\x{180a}\x{1944}\x{1945}'.
 139 '\x{2010}-\x{2027}\x{2030}-\x{2043}\x{2045}-\x{2051}\x{2053}\x{2054}\x{2057}'.
 140 '\x{207d}\x{207e}\x{208d}\x{208e}\x{2329}\x{232a}\x{23b4}-\x{23b6}\x{2768}-'.
 141 '\x{2775}\x{27e6}-\x{27eb}\x{2983}-\x{2998}\x{29d8}-\x{29db}\x{29fc}\x{29fd}'.
 142 '\x{3001}-\x{3003}\x{3008}-\x{3011}\x{3014}-\x{301f}\x{3030}\x{303d}\x{30a0}'.
 143 '\x{30fb}\x{fd3e}\x{fd3f}\x{fe30}-\x{fe52}\x{fe54}-\x{fe61}\x{fe63}\x{fe68}'.
 144 '\x{fe6a}\x{fe6b}\x{ff01}-\x{ff03}\x{ff05}-\x{ff0a}\x{ff0c}-\x{ff0f}\x{ff1a}'.
 145 '\x{ff1b}\x{ff1f}\x{ff20}\x{ff3b}-\x{ff3d}\x{ff3f}\x{ff5b}\x{ff5d}\x{ff5f}-'.
 146 '\x{ff65}');
 147
 148 /**
 149  * Matches all CJK characters that are candidates for auto-splitting
 150  * (Chinese, Japanese, Korean).
 151  * Contains kana and BMP ideographs.
 152  */
 153 define('PREG_CLASS_CJK', '\x{3041}-\x{30ff}\x{31f0}-\x{31ff}\x{3400}-\x{4db5}'.
 154 '\x{4e00}-\x{9fbb}\x{f900}-\x{fad9}');
 155
 156
 157 /**
 158  * This function process the text passed at input, extracting all the tokens
 159  * and scoring each one based in their number of ocurrences and relation with
 160  * some well-known html tags
 161  *
 162  * @param string  $text the text to be tokenised.
 163  * @param array   $stop_words array of utf-8 words than can be ignored in
 164  *                the text being processed. There are some cool lists of
 165  *                stop words at http://snowball.tartarus.org/
 166  * @param boolean $overlap_cjk option to split CJK text into some overlapping
 167  *                tokens is order to allow them to be searched. Useful to build
 168  *                indexes and search systems.
 169  * @param boolean $join_numbers option to join in one unique token sequences of numbers
 170  *                separated by puntuaction chars. Useful to build indexes and
 171  *                search systems.
 172  * @return array one sorted array of tokens, with tokens being the keys and scores in the values.
 173  */
 174 function tokenise_text($text, $stop_words = array(), $overlap_cjk = false, $join_numbers = false) {
 175
 176     // Multipliers for scores of words inside certain HTML tags.
 177     // Note: 'a' must be included for link ranking to work.
 178     $tags = array('h1' => 25,
 179                   'h2' => 18,
 180                   'h3' => 15,
 181                   'h4' => 12,
 182                   'h5' => 9,
 183                   'h6' => 6,
 184                   'u' => 3,
 185                   'b' => 3,
 186                   'i' => 3,
 187                   'strong' => 3,
 188                   'em' => 3,
 189                   'a' => 10);
 190
 191     // Strip off all ignored tags to speed up processing, but insert space before/after
 192     // them to keep word boundaries.
 193     $text = str_replace(array('<', '>'), array(' <', '> '), $text);
 194     $text = strip_tags($text, '<'. implode('><', array_keys($tags)) .'>');
 195
 196     // Split HTML tags from plain text.
 197     $split = preg_split('/\s*<([^>]+?)>\s*/', $text, -1, PREG_SPLIT_DELIM_CAPTURE);
 198     // Note: PHP ensures the array consists of alternating delimiters and literals
 199     // and begins and ends with a literal (inserting $null as required).
 200
 201     $tag = FALSE; // Odd/even counter. Tag or no tag.
 202     $score = 1; // Starting score per word
 203     $accum = ' '; // Accumulator for cleaned up data
 204     $tagstack = array(); // Stack with open tags
 205     $tagwords = 0; // Counter for consecutive words
 206     $focus = 1; // Focus state
 207
 208     $results = array(0 => array()); // Accumulator for words for index
 209
 210     foreach ($split as $value) {
 211         if ($tag) {
 212             // Increase or decrease score per word based on tag
 213             list($tagname) = explode(' ', $value, 2);
 214             $tagname = core_text::strtolower($tagname);
 215             // Closing or opening tag?
 216             if ($tagname[0] == '/') {
 217                 $tagname = substr($tagname, 1);
 218                 // If we encounter unexpected tags, reset score to avoid incorrect boosting.
 219                 if (!count($tagstack) || $tagstack[0] != $tagname) {
 220                     $tagstack = array();
 221                     $score = 1;
 222                 }
 223                 else {
 224                     // Remove from tag stack and decrement score
 225                     $score = max(1, $score - $tags[array_shift($tagstack)]);
 226                 }
 227             }
 228             else {
 229                 if (isset($tagstack[0]) && $tagstack[0] == $tagname) {
 230                     // None of the tags we look for make sense when nested identically.
 231                     // If they are, it's probably broken HTML.
 232                     $tagstack = array();
 233                     $score = 1;
 234                 }
 235                 else {
 236                     // Add to open tag stack and increment score
 237                     array_unshift($tagstack, $tagname);
 238                     $score += $tags[$tagname];
 239                 }
 240             }
 241             // A tag change occurred, reset counter.
 242             $tagwords = 0;
 243         }
 244         else {
 245             // Note: use of PREG_SPLIT_DELIM_CAPTURE above will introduce empty values
 246             if ($value != '') {
 247                 $words = tokenise_split($value, $stop_words, $overlap_cjk, $join_numbers);
 248                 foreach ($words as $word) {
 249                     // Add word to accumulator
 250                     $accum .= $word .' ';
 251                     $num = is_numeric($word);
 252                     // Check word length
 253                     if ($num || core_text::strlen($word) >= MINIMUM_WORD_SIZE) {
 254                         // Normalize numbers
 255                         if ($num && $join_numbers) {
 256                             $word = (int)ltrim($word, '-0');
 257                         }
 258
 259                         if (!isset($results[0][$word])) {
 260                             $results[0][$word] = 0;
 261                         }
 262
 263                         $results[0][$word] += $score * $focus;
 264                         // Focus is a decaying value in terms of the amount of unique words up to this point.
 265                         // From 100 words and more, it decays, to e.g. 0.5 at 500 words and 0.3 at 1000 words.
 266                         $focus = min(1, .01 + 3.5 / (2 + count($results[0]) * .015));
 267                     }
 268                     $tagwords++;
 269                     // Too many words inside a single tag probably mean a tag was accidentally left open.
 270                     if (count($tagstack) && $tagwords >= 15) {
 271                         $tagstack = array();
 272                         $score = 1;
 273                     }
 274                 }
 275             }
 276         }
 277         $tag = !$tag;
 278     }
 279
 280     $res = array();
 281
 282     if (isset($results[0]) && count($results[0]) > 0) {
 283         $res = $results[0];
 284         arsort($res, SORT_NUMERIC);
 285     }
 286
 287     return $res;
 288 }
 289
 290 ///
 291 /// Some helper functions (should be considered private)
 292 ///
 293
 294 /**
 295  * Splits a string into tokens
 296  */
 297 function tokenise_split($text, $stop_words, $overlap_cjk, $join_numbers) {
 298     static $last = NULL;
 299     static $lastsplit = NULL;
 300
 301     if ($last == $text) {
 302         return $lastsplit;
 303     }
 304     // Process words
 305     $text = tokenise_simplify($text, $overlap_cjk, $join_numbers);
 306     $words = explode(' ', $text);
 307     // Limit word length to 50
 308     array_walk($words, 'tokenise_truncate_word');
 309
 310     // We have stop words, apply them
 311     if (is_array($stop_words) && !empty($stop_words)) {
 312         // Normalise them
 313         $simp_stop_words = explode(' ', tokenise_simplify(implode(' ', $stop_words), $overlap_cjk, $join_numbers));
 314         // Extract from the list
 315         $words = array_diff($words, $simp_stop_words);
 316     }
 317     // Save last keyword result
 318     $last = $text;
 319     $lastsplit = $words;
 320
 321     return $words;
 322 }
 323
 324 /**
 325  * Simplifies a string according to indexing rules.
 326  */
 327 function tokenise_simplify($text, $overlap_cjk, $join_numbers) {
 328
 329     // Decode entities to UTF-8
 330     $text = core_text::entities_to_utf8($text, true);
 331
 332     // Lowercase
 333     $text = core_text::strtolower($text);
 334
 335     // Simple CJK handling
 336     if ($overlap_cjk) {
 337         $text = preg_replace_callback('/['. PREG_CLASS_CJK .']+/u', 'tokenise_expand_cjk', $text);
 338     }
 339
 340     // To improve searching for numerical data such as dates, IP addresses
 341     // or version numbers, we consider a group of numerical characters
 342     // separated only by punctuation characters to be one piece.
 343     // This also means that searching for e.g. '20/03/1984' also returns
 344     // results with '20-03-1984' in them.
 345     // Readable regexp: ([number]+)[punctuation]+(?=[number])
 346     if ($join_numbers) {
 347         $text = preg_replace('/(['. PREG_CLASS_NUMBERS .']+)['. PREG_CLASS_PUNCTUATION .']+(?=['. PREG_CLASS_NUMBERS .'])/u', '\1', $text);
 348     } else {
 349     // Keep all the detected numbers+punctuation in a safe place in order to restore them later
 350         preg_match_all('/['. PREG_CLASS_NUMBERS .']+['. PREG_CLASS_PUNCTUATION . PREG_CLASS_NUMBERS .']+/u', $text, $foundseqs);
 351         $storedseqs = array();
 352         foreach (array_unique($foundseqs[0]) as $ntkey => $value) {
 353             $prefix = (string)(count($storedseqs) + 1);
 354             $storedseqs[START_DELIM.$prefix.CENTER_DELIM.$ntkey.END_DELIM] = $value;
 355         }
 356         if (!empty($storedseqs)) {
 357             $text = str_replace($storedseqs, array_keys($storedseqs), $text);
 358         }
 359     }
 360
 361     // The dot, underscore and dash are simply removed. This allows meaningful
 362     // search behaviour with acronyms and URLs.
 363     $text = preg_replace('/[._-]+/', '', $text);
 364
 365     // With the exception of the rules above, we consider all punctuation,
 366     // marks, spacers, etc, to be a word boundary.
 367     $text = preg_replace('/['. PREG_CLASS_SEARCH_EXCLUDE .']+/u', ' ', $text);
 368
 369     // Restore, if not joining numbers, recover the original strings
 370     if (!$join_numbers) {
 371         if (!empty($storedseqs)) {
 372             $text = str_replace(array_keys($storedseqs), $storedseqs, $text);
 373         }
 374     }
 375
 376     return $text;
 377 }
 378
 379 /**
 380  * Basic CJK tokeniser. Simply splits a string into consecutive, overlapping
 381  * sequences of characters (MINIMUM_WORD_SIZE long).
 382  */
 383 function tokenise_expand_cjk($matches) {
 384
 385     $str = $matches[0];
 386     $l = core_text::strlen($str);
 387     // Passthrough short words
 388     if ($l <= MINIMUM_WORD_SIZE) {
 389         return ' '. $str .' ';
 390     }
 391     $tokens = ' ';
 392     // FIFO queue of characters
 393     $chars = array();
 394     // Begin loop
 395     for ($i = 0; $i < $l; ++$i) {
 396         // Grab next character
 397         $current = core_text::substr($str, 0, 1);
 398         $str = substr($str, strlen($current));
 399         $chars[] = $current;
 400         if ($i >= MINIMUM_WORD_SIZE - 1) {
 401             $tokens .= implode('', $chars) .' ';
 402             array_shift($chars);
 403         }
 404     }
 405     return $tokens;
 406 }
 407
 408 /**
 409  * Helper function for array_walk in search_index_split.
 410  * Truncates one string (token) to MAXIMUM_WORD_SIZE
 411  */
 412 function tokenise_truncate_word(&$text) {
 413
 414     $text = core_text::substr($text, 0, MAXIMUM_WORD_SIZE);
 415 }