<?php
/**
* Detects the language of a given piece of text.
*
* Attempts to detect the language of a sample of text by correlating ranked
* 3-gram frequencies to a table of 3-gram frequencies of known languages.
*
* Implements a version of a technique originally proposed by Cavnar & Trenkle
* (1994): "N-Gram-Based Text Categorization"
*
* PHP versions 4 and 5
*
* @category Text
* @package Text_LanguageDetect
* @author Nicholas Pisarro <infinityminusnine+pear@gmail.com>
* @copyright 2005-2006 Nicholas Pisarro
* @license http://www.debian.org/misc/bsd.license BSD
* @version CVS: $Id: LanguageDetect.php,v 1.20 2008/07/01 02:09:15 taak Exp $
* @link http://pear.php.net/package/Text_LanguageDetect/
* @link http://langdetect.blogspot.com/
*/
//require_once 'PEAR.php';
require_once 'Parser.php';
/**
* Language detection class
*
* Requires the langauge model database (lang.dat) that should have
* accompanied this class definition in order to be instantiated.
*
* Example usage:
*
* <code>
* require_once 'Text/LanguageDetect.php';
*
* $l = new Text_LanguageDetect;
*
* $stdin = fopen('php://stdin', 'r');
*
* echo "Supported languages:\n";
*
* $langs = $l->getLanguages();
* if (PEAR::isError($langs)) {
* die($langs->getMessage());
* }
*
* sort($langs);
* echo join(', ', $langs);
*
* while ($line = fgets($stdin)) {
* print_r($l->detect($line, 4));
* }
* </code>
*
* @category Text
* @package Text_LanguageDetect
* @author Nicholas Pisarro <infinityminusnine+pear@gmail.com>
* @copyright 2005 Nicholas Pisarro
* @license http://www.debian.org/misc/bsd.license BSD
* @version Release: @package_version@
* @todo allow users to generate their own language models
*/
class Text_LanguageDetect
{
/**
* The filename that stores the trigram data for the detector
*
* If this value starts with a slash (/) or a dot (.) the value of
* $this->_data_dir will be ignored
*
* @var string
* @access private
*/
var $_db_filename = './lang.dat';
/**
* The filename that stores the unicode block definitions
*
* If this value starts with a slash (/) or a dot (.) the value of
* $this->_data_dir will be ignored
*
* @var string
* @access private
*/
var $_unicode_db_filename = './unicode_blocks.dat';
/**
* The data directory
*
* Should be set by PEAR installer
*
* @var string
* @access private
*/
var $_data_dir = '@data_dir@';
/**
* The trigram data for comparison
*
* Will be loaded on start from $this->_db_filename
*
* May be set to a PEAR_Error object if there is an error during its
* initialization
*
* @var array
* @access private
*/
var $_lang_db = array();
/**
* stores the map of the trigram data to unicode characters
*
* @access private
* @var array
*/
var $_unicode_map;
/**
* The size of the trigram data arrays
*
* @var int
* @access private
*/
var $_threshold = 300;
/**
* the maximum possible score.
*
* needed for score normalization. Different depending on the
* perl compatibility setting
*
* @access private
* @var int
* @see setPerlCompatible()
*/
var $_max_score = 0;
/**
* Whether or not to simulate perl's Language::Guess exactly
*
* @access private
* @var bool
* @see setPerlCompatible()
*/
var $_perl_compatible = false;
/**
* Whether to use the unicode block detection to speed up processing
*
* @access private
* @var bool
*/
var $_use_unicode_narrowing = true;
/**
* stores the result of the clustering operation
*
* @access private
* @var array
* @see clusterLanguages()
*/
var $_clusters;
/**
* Constructor
*
* Will attempt to load the language database. If it fails, you will get
* a PEAR_Error object returned when you try to use detect()
*
*/
function Text_LanguageDetect($db=null, $unicode_db=null)
{
if (isset($db)) $this->_db_filename = $db;
if (isset($unicode_db)) $this->_unicode_db_filename = $unicode_db;
$data = $this->_readdb($this->_db_filename);
$this->_lang_db = $data['trigram'];
if (isset($data['trigram-unicodemap'])) {
$this->_unicode_map = $data['trigram-unicodemap'];
}
// Not yet implemented:
if (isset($data['trigram-clusters'])) {
$this->_clusters = $data['trigram-clusters'];
}
}
/**
* Returns the path to the location of the database
*
* @access private
* @return string expected path to the language model database
*/
function _get_data_loc($fname)
{
return $fname;
}
/**
* Loads the language trigram database from filename
*
* Trigram datbase should be a serialize()'d array
*
* @access private
* @param string $fname the filename where the data is stored
* @return array the language model data
* @throws PEAR_Error
*/
function _readdb($fname)
{
// finds the correct data dir
$fname = $this->_get_data_loc($fname);
// input check
if (!file_exists($fname)) {
throw new Exception('Language database does not exist.');
} elseif (!is_readable($fname)) {
throw new Exception('Language database is not readable.');
}
if (function_exists('file_get_contents')) {
return unserialize(file_get_contents($fname));
} else {
// if you don't have file_get_contents(),
// then this is the next fastest way
ob_start();
readfile($fname);
$contents = ob_get_contents();
ob_end_clean();
return unserialize($contents);
}
}
/**
* Checks if this object is ready to detect languages
*
* @access private
* @param mixed &$err error object to be returned by reference, if any
* @return bool true if no errors
*/
function _setup_ok(&$err)
{
if (!is_array($this->_lang_db)) {
if (ini_get('magic_quotes_runtime')) {
throw new Exception('Error loading database. Try turning magic_quotes_runtime off.');
} else {
throw new Exception('Language database is not an array.');
}
return false;
} elseif (empty($this->_lang_db)) {
throw new Exception('Language database has no elements.');
return false;
} else {
return true;
}
}
/**
* Omits languages
*
* Pass this function the name of or an array of names of
* languages that you don't want considered
*
* If you're only expecting a limited set of languages, this can greatly
* speed up processing
*
* @access public
* @param mixed $omit_list language name or array of names to omit
* @param bool $include_only if true will include (rather than
* exclude) only those in the list
* @return int number of languages successfully deleted
* @throws PEAR_Error
*/
function omitLanguages($omit_list, $include_only = false)
{
// setup check
if (!$this->_setup_ok($err)) {
return $err;
}
$deleted = 0;
// deleting the given languages
if (!$include_only) {
if (!is_array($omit_list)) {
$omit_list = strtolower($omit_list); // case desensitize
if (isset($this->_lang_db[$omit_list])) {
unset($this->_lang_db[$omit_list]);
$deleted++;
}
} else {
foreach ($omit_list as $omit_lang) {
if (isset($this->_lang_db[$omit_lang])) {
unset($this->_lang_db[$omit_lang]);
$deleted++;
}
}
}
// deleting all except the given languages
} else {
if (!is_array($omit_list)) {
$omit_list = array($omit_list);
}
// case desensitize
foreach ($omit_list as $key => $omit_lang) {
$omit_list[$key] = strtolower($omit_lang);
}
foreach (array_keys($this->_lang_db) as $lang) {
if (!in_array($lang, $omit_list)) {
unset($this->_lang_db[$lang]);
$deleted++;
}
}
}
// reset the cluster cache if the number of languages changes
// this will then have to be recalculated
if (isset($this->_clusters) && $deleted > 0) {
unset($this->_clusters);
}
return $deleted;
}
/**
* Returns the number of languages that this object can detect
*
* @access public
* @return int the number of languages
* @throws PEAR_Error
*/
function getLanguageCount()
{
if (!$this->_setup_ok($err)) {
return $err;
} else {
return count($this->_lang_db);
}
}
/**
* Returns true if a given language exists
*
* If passed an array of names, will return true only if all exist
*
* @access public
* @param mixed $lang language name or array of language names
* @return bool true if language model exists
* @throws PEAR_Error
*/
function languageExists($lang)
{
if (!$this->_setup_ok($err)) {
return $err;
} else {
// string
if (is_string($lang)) {
return isset($this->_lang_db[strtolower($lang)]);
// array
} elseif (is_array($lang)) {
foreach ($lang as $test_lang) {
if (!isset($this->_lang_db[strtolower($test_lang)])) {
return false;
}
}
return true;
// other (error)
} else {
throw new Exception('Unknown type passed to languageExists()');
}
}
}
/**
* Returns the list of detectable languages
*
* @access public
* @return array the names of the languages known to this object
* @throws PEAR_Error
*/
function getLanguages()
{
if (!$this->_setup_ok($err)) {
return $err;
} else {
return array_keys($this->_lang_db);
}
}
/**
* Make this object behave like Language::Guess
*
* @access public
* @param bool $setting false to turn off perl compatibility
*/
function setPerlCompatible($setting = true)
{
if (is_bool($setting)) { // input check
$this->_perl_compatible = $setting;
if ($setting == true) {
$this->_max_score = $this->_threshold;
} else {
$this->_max_score = 0;
}
}
}
/**
* Whether to use unicode block ranges in detection
*
* Should speed up most detections if turned on (detault is on). In some
* circumstances it may be slower, such as for large text samples (> 10K)
* in languages that use latin scripts. In other cases it should speed up
* detection noticeably.
*
* @access public
* @param bool $setting false to turn off
*/
function useUnicodeBlocks($setting = true)
{
if (is_bool($setting)) {
$this->_use_unicode_narrowing = $setting;
}
}
/**
* Converts a piece of text into trigrams
*
* Superceded by the Text_LanguageDetect_Parser class
*
* @access private
* @param string $text text to convert
* @return array array of trigram frequencies
*/
function _trigram($text)
{
$s = new Text_LanguageDetect_Parser($text, $this->_db_filename, $this->_unicode_db_filename);
$s->prepareTrigram();
$s->prepareUnicode(false);
$s->setPadStart(!$this->_perl_compatible);
$s->analyze();
return $s->getTrigramFreqs();
}
/**
* Converts a set of trigrams from frequencies to ranks
*
* Thresholds (cuts off) the list at $this->_threshold
*
* @access protected
* @param array $arr array of trgram
* @return array ranks of trigrams
*/
function _arr_rank(&$arr)
{
// sorts alphabetically first as a standard way of breaking rank ties
$this->_bub_sort($arr);
// below might also work, but seemed to introduce errors in testing
//ksort($arr);
//asort($arr);
$rank = array();
$i = 0;
foreach ($arr as $key => $value) {
$rank[$key] = $i++;
// cut off at a standard threshold
if ($i >= $this->_threshold) {
break;
}
}
return $rank;
}
/**
* Sorts an array by value breaking ties alphabetically
*
* @access private
* @param array &$arr the array to sort
*/
function _bub_sort(&$arr)
{
// should do the same as this perl statement:
// sort { $trigrams{$b} == $trigrams{$a} ? $a cmp $b : $trigrams{$b} <=> $trigrams{$a} }
// needs to sort by both key and value at once
// using the key to break ties for the value
// converts array into an array of arrays of each key and value
// may be a better way of doing this
$combined = array();
foreach ($arr as $key => $value) {
$combined[] = array($key, $value);
}
usort($combined, array($this, '_sort_func'));
$replacement = array();
foreach ($combined as $key => $value) {
list($new_key, $new_value) = $value;
$replacement[$new_key] = $new_value;
}
$arr = $replacement;
}
/**
* Sort function used by bubble sort
*
* Callback function for usort().
*
* @access private
* @param array first param passed by usort()
* @param array second param passed by usort()
* @return int 1 if $a is greater, -1 if not
* @see _bub_sort()
*/
function _sort_func($a, $b)
{
// each is actually a key/value pair, so that it can compare using both
list($a_key, $a_value) = $a;
list($b_key, $b_value) = $b;
// if the values are the same, break ties using the key
if ($a_value == $b_value) {
return strcmp($a_key, $b_key);
// if not, just sort normally
} else {
if ($a_value > $b_value) {
return -1;
} else {
return 1;
}
}
// 0 should not be possible because keys must be unique
}
/**
* Calculates a linear rank-order distance statistic between two sets of
* ranked trigrams
*
* Sums the differences in rank for each trigram. If the trigram does not
* appear in both, consider it a difference of $this->_threshold.
*
* This distance measure was proposed by Cavnar & Trenkle (1994). Despite
* its simplicity it has been shown to be highly accurate for language
* identification tasks.
*
* @access private
* @param array $arr1 the reference set of trigram ranks
* @param array $arr2 the target set of trigram ranks
* @return int the sum of the differences between the ranks of
* the two trigram sets
*/
function _distance(&$arr1, &$arr2)
{
$sumdist = 0;
foreach ($arr2 as $key => $value) {
if (isset($arr1[$key])) {
$distance = abs($value - $arr1[$key]);
} else {
// $this->_threshold sets the maximum possible distance value
// for any one pair of trigrams
$distance = $this->_threshold;
}
$sumdist += $distance;
}
return $sumdist;
// todo: there are other distance statistics to try, e.g. relative
// entropy, but they're probably more costly to compute
}
/**
* Normalizes the score returned by _distance()
*
* Different if perl compatible or not
*
* @access private
* @param int $score the score from _distance()
* @param int $base_count the number of trigrams being considered
* @return float the normalized score
* @see _distance()
*/
function _normalize_score($score, $base_count = null)
{
if ($base_count === null) {
$base_count = $this->_threshold;
}
if (!$this->_perl_compatible) {
return 1 - ($score / $base_count / $this->_threshold);
} else {
return floor($score / $base_count);
}
}
/**
* Detects the closeness of a sample of text to the known languages
*
* Calculates the statistical difference between the text and
* the trigrams for each language, normalizes the score then
* returns results for all languages in sorted order
*
* If perl compatible, the score is 300-0, 0 being most similar.
* Otherwise, it's 0-1 with 1 being most similar.
*
* The $sample text should be at least a few sentences in length;
* should be ascii-7 or utf8 encoded, if another and the mbstring extension
* is present it will try to detect and convert. However, experience has
* shown that mb_detect_encoding() *does not work very well* with at least
* some types of encoding.
*
* @access public
* @param string $sample a sample of text to compare.
* @param int $limit if specified, return an array of the most likely
* $limit languages and their scores.
* @return mixed sorted array of language scores, blank array if no
* useable text was found, or PEAR_Error if error
* with the object setup
* @see _distance()
* @throws PEAR_Error
*/
function detect($sample, $limit = 0)
{
if (!$this->_setup_ok($err)) {
return $err;
}
// input check
if (!Text_LanguageDetect_Parser::validateString($sample)) {
return array();
}
// check char encoding
// (only if mbstring extension is compiled and PHP > 4.0.6)
if (function_exists('mb_detect_encoding')
&& function_exists('mb_convert_encoding')) {
// mb_detect_encoding isn't very reliable, to say the least
// detection should still work with a sufficient sample of ascii characters
$encoding = mb_detect_encoding($sample);
// mb_detect_encoding() will return FALSE if detection fails
// don't attempt conversion if that's the case
if ($encoding != 'ASCII' && $encoding != 'UTF-8' && $encoding !== false) {
if (function_exists('mb_list_encodings')) {
// verify the encoding exists in mb_list_encodings
if (in_array($encoding, mb_list_encodings())) {
$sample = mb_convert_encoding($sample, 'UTF-8', $encoding);
}
// if the previous condition failed:
// somehow we detected an encoding that also we don't support
} else {
// php 4 doesnt have mb_list_encodings()
// so attempt with error suppression
$sample = @mb_convert_encoding($sample, 'UTF-8', $encoding);
}
}
}
$sample_obj = new Text_LanguageDetect_Parser($sample, $this->_db_filename, $this->_unicode_db_filename);
$sample_obj->prepareTrigram();
if ($this->_use_unicode_narrowing) {
$sample_obj->prepareUnicode();
}
$sample_obj->setPadStart(!$this->_perl_compatible);
$sample_obj->analyze();
$trigram_freqs =& $sample_obj->getTrigramRanks();
$trigram_count = count($trigram_freqs);
if ($trigram_count == 0) {
return array();
}
$scores = array();
// use unicode block detection to narrow down the possibilities
if ($this->_use_unicode_narrowing) {
$blocks =& $sample_obj->getUnicodeBlocks();
if (is_array($blocks)) {
$present_blocks = array_keys($blocks);
} else {
throw new Exception('Error during block detection');
}
$possible_langs = array();
foreach ($present_blocks as $blockname) {
if (isset($this->_unicode_map[$blockname])) {
$possible_langs = array_merge(
$possible_langs,
array_keys($this->_unicode_map[$blockname])
);
// todo: faster way to do this?
}
}
// could also try an intersect operation rather than a union
// in other words, choose languages whose trigrams contain
// ALL of the unicode blocks found in this sample
// would improve speed but would be completely thrown off by an
// unexpected character, like an umlaut appearing in english text
$possible_langs = array_intersect(
array_keys($this->_lang_db),
array_unique($possible_langs)
);
// needs to intersect it with the keys of _lang_db in case
// languages have been omitted
// or just try 'em all
} else {
$possible_langs = array_keys($this->_lang_db);
}
foreach ($possible_langs as $lang) {
$scores[$lang] =
$this->_normalize_score(
$this->_distance($this->_lang_db[$lang], $trigram_freqs),
$trigram_count);
}
unset($sample_obj);
if ($this->_perl_compatible) {
asort($scores);
} else {
arsort($scores);
}
// todo: drop languages with a score of $this->_max_score?
// limit the number of returned scores
if ($limit && is_numeric($limit)) {
$limited_scores = array();
$i = 0;
foreach ($scores as $key => $value) {
if ($i++ >= $limit) {
break;
}
$limited_scores[$key] = $value;
}
return $limited_scores;
} else {
return $scores;
}
}
/**
* Returns only the most similar language to the text sample
*
* Calls $this->detect() and returns only the top result
*
* @access public
* @param string $sample text to detect the language of
* @return string the name of the most likely language
* or null if no language is similar
* @see detect()
* @throws PEAR_Error
*/
function detectSimple($sample)
{
$scores = $this->detect($sample, 1);
// if top language has the maximum possible score,
// then the top score will have been picked at random
if ( !is_array($scores)
|| empty($scores)
|| current($scores) == $this->_max_score) {
return null;
} else {
return ucfirst(key($scores));
}
}
/**
* Returns an array containing the most similar language and a confidence
* rating
*
* Confidence is a simple measure calculated from the similarity score
* minus the similarity score from the next most similar language
* divided by the highest possible score. Languages that have closely
* related cousins (e.g. Norwegian and Danish) should generally have lower
* confidence scores.
*
* The similarity score answers the question "How likely is the text the
* returned language regardless of the other languages considered?" The
* confidence score is one way of answering the question "how likely is the
* text the detected language relative to the rest of the language model
* set?"
*
* To see how similar languages are a priori, see languageSimilarity()
*
* @access public
* @param string $sample text for which language will be detected
* @return array most similar language, score and confidence rating
* or null if no language is similar
* @see detect()
* @throws PEAR_Error
*/
function detectConfidence($sample)
{
$scores = $this->detect($sample, 2);
// if most similar language has the max score, it
// will have been picked at random
if ( !is_array($scores)
|| empty($scores)
|| current($scores) == $this->_max_score) {
return null;
}
$arr['language'] = ucfirst(key($scores));
$arr['similarity'] = current($scores);
if (next($scores) !== false) { // if false then no next element
// the goal is to return a higher value if the distance between
// the similarity of the first score and the second score is high
if ($this->_perl_compatible) {
$arr['confidence'] =
(current($scores) - $arr['similarity']) / $this->_max_score;
} else {
$arr['confidence'] = $arr['similarity'] - current($scores);
}
} else {
$arr['confidence'] = null;
}
return $arr;
}
/**
* Returns the distribution of unicode blocks in a given utf8 string
*
* For the block name of a single char, use unicodeBlockName()
*
* @access public
* @param string $str input string. Must be ascii or utf8
* @param bool $skip_symbols if true, skip ascii digits, symbols and
* non-printing characters. Includes spaces,
* newlines and common punctutation characters.
* @return array
* @throws PEAR_Error
*/
function detectUnicodeBlocks($str, $skip_symbols)
{
// input check
if (!is_bool($skip_symbols)) {
throw new Exception('Second parameter must be boolean');
}
if (!is_string($str)) {
throw new Exception('First parameter was not a string');
}
$sample_obj = new Text_LanguageDetect_Parser($str, $this->_db_filename, $this->_unicode_db_filename);
$sample_obj->prepareUnicode();
$sample_obj->prepareTrigram(false);
$sample_obj->setUnicodeSkipSymbols($skip_symbols);
$sample_obj->analyze();
$blocks =& $sample_obj->getUnicodeBlocks();
unset($sample_obj);
return $blocks;
}
/**
* Returns the block name for a given unicode value
*
* If passed a string, will assume it is being passed a UTF8-formatted
* character and will automatically convert. Otherwise it will assume it
* is being passed a numeric unicode value.
*
* Make sure input is of the correct type!
*
* @access public
* @param mixed $unicode unicode value or utf8 char
* @return mixed the block name string or false if not found
* @throws PEAR_Error
*/
function unicodeBlockName($unicode) {
if (is_string($unicode)) {
// assume it is being passed a utf8 char, so convert it
// input check
if ($this->utf8strlen($unicode) > 1) {
throw new Exception('Pass this function only a single char');
}
$unicode = $this->_utf8char2unicode($unicode);
if ($unicode == -1) {
throw new Exception('Malformatted char');
}
// input check
} elseif (!is_int($unicode)) {
throw new Exception('Input must be of type string or int.');
}
$blocks =& $this->_read_unicode_block_db();
$result = $this->_unicode_block_name($unicode, $blocks);
if ($result == -1) {
return false;
} else {
return $result[2];
}
}
/**
* Searches the unicode block database
*
* Returns the block name for a given unicode value. unicodeBlockName() is
* the public interface for this function, which does input checks which
* this function omits for speed.
*
* @access protected
* @param int $unicode the unicode value
* @param array &$blocks the block database
* @param int $block_count the number of defined blocks in the database
* @see unicodeBlockName()
*/
function _unicode_block_name($unicode, &$blocks, $block_count = -1) {
// for a reference, see
// http://www.unicode.org/Public/UNIDATA/Blocks.txt
// assume that ascii characters are the most common
// so try it first for efficiency
if ($unicode <= $blocks[0][1]) {
return $blocks[0];
}
// the optional $block_count param is for efficiency
// so we this function doesn't have to run count() every time
if ($block_count != -1) {
$high = $block_count - 1;
} else {
$high = count($blocks) - 1;
}
$low = 1; // start with 1 because ascii was 0
// your average binary search algorithm
while ($low <= $high) {
$mid = floor(($low + $high) / 2);
// if it's lower than the lower bound
if ($unicode < $blocks[$mid][0]) {
$high = $mid - 1;
// if it's higher than the upper bound
} elseif ($unicode > $blocks[$mid][1]) {
$low = $mid + 1;
// found it
} else {
return $blocks[$mid];
}
}
// failed to find the block
return -1;
// todo: differentiate when it's out of range or when it falls
// into an unassigned range?
}
/**
* Brings up the unicode block database
*
* @access protected
* @return array the database of unicode block definitions
* @throws PEAR_Error
*/
function &_read_unicode_block_db() {
// since the unicode definitions are always going to be the same,
// might as well share the memory for the db with all other instances
// of this class
static $data;
if (!isset($data)) {
$data = $this->_readdb($this->_unicode_db_filename);
}
return $data;
}
/**
* Calculate the similarities between the language models
*
* Use this function to see how similar languages are to each other.
*
* If passed 2 language names, will return just those languages compared.
* If passed 1 language name, will return that language compared to
* all others.
* If passed none, will return an array of every language model compared
* to every other one.
*
* @access public
* @param string $lang1 the name of the first language to be compared
* @param string $lang2 the name of the second language to be compared
* @return array scores of every language compared
* or the score of just the provided languages
* or null if one of the supplied languages does not exist
* @throws PEAR_Error
*/
function languageSimilarity($lang1 = null, $lang2 = null)
{
if (!$this->_setup_ok($err)) {
return $err;
}
if ($lang1 != null) {
$lang1 = strtolower($lang1);
// check if language model exists
if (!isset($this->_lang_db[$lang1])) {
return null;
}
if ($lang2 != null) {
// can't only set the second param
if ($lang1 == null) {
return null;
// check if language model exists
} elseif (!isset($this->_lang_db[$lang2])) {
return null;
}
$lang2 = strtolower($lang2);
// compare just these two languages
return $this->_normalize_score(
$this->_distance(
$this->_lang_db[$lang1],
$this->_lang_db[$lang2]
)
);
// compare just $lang1 to all languages
} else {
$return_arr = array();
foreach ($this->_lang_db as $key => $value) {
if ($key != $lang1) { // don't compare a language to itself
$return_arr[$key] = $this->_normalize_score(
$this->_distance($this->_lang_db[$lang1], $value));
}
}
asort($return_arr);
return $return_arr;
}
// compare all languages to each other
} else {
$return_arr = array();
foreach (array_keys($this->_lang_db) as $lang1) {
foreach (array_keys($this->_lang_db) as $lang2) {
// skip comparing languages to themselves
if ($lang1 != $lang2) {
// don't re-calculate what's already been done
if (isset($return_arr[$lang2][$lang1])) {
$return_arr[$lang1][$lang2] =
$return_arr[$lang2][$lang1];
// calculate
} else {
$return_arr[$lang1][$lang2] =
$this->_normalize_score(
$this->_distance(
$this->_lang_db[$lang1],
$this->_lang_db[$lang2]
)
);
}
}
}
}
return $return_arr;
}
}
/**
* Cluster known languages according to languageSimilarity()
*
* WARNING: this method is EXPERIMENTAL. It is not recommended for common
* use, and it may disappear or its functionality may change in future
* releases without notice.
*
* Uses a nearest neighbor technique to generate the maximum possible
* number of dendograms from the similarity data.
*
* @access public
* @return array language cluster data
* @throws PEAR_Error
* @see languageSimilarity()
* @deprecated this function will eventually be removed and placed into
* the model generation class
*/
function clusterLanguages()
{
// todo: set the maximum number of clusters
// setup check
if (!$this->_setup_ok($err)) {
return $err;
}
// return cached result, if any
if (isset($this->_clusters)) {
return $this->_clusters;
}
$langs = array_keys($this->_lang_db);
$arr = $this->languageSimilarity();
sort($langs);
foreach ($langs as $lang) {
if (!isset($this->_lang_db[$lang])) {
throw new Exception("missing $lang!\n");
}
}
// http://www.psychstat.missouristate.edu/multibook/mlt04m.html
foreach ($langs as $old_key => $lang1) {
$langs[$lang1] = $lang1;
unset($langs[$old_key]);
}
$i = 0;
while (count($langs) > 2 && $i++ < 200) {
$highest_score = -1;
$highest_key1 = '';
$highest_key2 = '';
foreach ($langs as $lang1) {
foreach ($langs as $lang2) {
if ( $lang1 != $lang2
&& $arr[$lang1][$lang2] > $highest_score) {
$highest_score = $arr[$lang1][$lang2];
$highest_key1 = $lang1;
$highest_key2 = $lang2;
}
}
}
if (!$highest_key1) {
// should not ever happen
throw new Exception("no highest key? (step: $i)");
}
if ($highest_score == 0) {
// languages are perfectly dissimilar
break;
}
// $highest_key1 and $highest_key2 are most similar
$sum1 = array_sum($arr[$highest_key1]);
$sum2 = array_sum($arr[$highest_key2]);
// use the score for the one that is most similar to the rest of
// the field as the score for the group
// todo: could try averaging or "centroid" method instead
// seems like that might make more sense
// actually nearest neighbor may be better for binary searching
// for "Complete Linkage"/"furthest neighbor"
// sign should be <
// for "Single Linkage"/"nearest neighbor" method
// should should be >
// results seem to be pretty much the same with either method
// figure out which to delete and which to replace
if ($sum1 > $sum2) {
$replaceme = $highest_key1;
$deleteme = $highest_key2;
} else {
$replaceme = $highest_key2;
$deleteme = $highest_key1;
}
$newkey = $replaceme . ':' . $deleteme;
// $replaceme is most similar to remaining languages
// replace $replaceme with '$newkey', deleting $deleteme
// keep a record of which fork is really which language
$really_lang = $replaceme;
while (isset($really_map[$really_lang])) {
$really_lang = $really_map[$really_lang];
}
$really_map[$newkey] = $really_lang;
// replace the best fitting key, delete the other
foreach ($arr as $key1 => $arr2) {
foreach ($arr2 as $key2 => $value2) {
if ($key2 == $replaceme) {
$arr[$key1][$newkey] = $arr[$key1][$key2];
unset($arr[$key1][$key2]);
// replacing $arr[$key1][$key2] with $arr[$key1][$newkey]
}
if ($key1 == $replaceme) {
$arr[$newkey][$key2] = $arr[$key1][$key2];
unset($arr[$key1][$key2]);
// replacing $arr[$key1][$key2] with $arr[$newkey][$key2]
}
if ($key1 == $deleteme || $key2 == $deleteme) {
// deleting $arr[$key1][$key2]
unset($arr[$key1][$key2]);
}
}
}
unset($langs[$highest_key1]);
unset($langs[$highest_key2]);
$langs[$newkey] = $newkey;
// some of these may be overkill
$result_data[$newkey] = array(
'newkey' => $newkey,
'count' => $i,
'diff' => abs($sum1 - $sum2),
'score' => $highest_score,
'bestfit' => $replaceme,
'otherfit' => $deleteme,
'really' => $really_lang,
);
}
$return_val = array(
'open_forks' => $langs,
// the top level of clusters
// clusters that are mutually exclusive
// or specified by a specific maximum
'fork_data' => $result_data,
// data for each split
'name_map' => $really_map,
// which cluster is really which language
// using the nearest neighbor technique, the cluster
// inherits all of the properties of its most-similar member
// this keeps track
);
// saves the result in the object
$this->_clusters = $return_val;
return $return_val;
}
/**
* Perform an intelligent detection based on clusterLanguages()
*
* WARNING: this method is EXPERIMENTAL. It is not recommended for common
* use, and it may disappear or its functionality may change in future
* releases without notice.
*
* This compares the sample text to top the top level of clusters. If the
* sample is similar to the cluster it will drop down and compare it to the
* languages in the cluster, and so on until it hits a leaf node.
*
* this should find the language in considerably fewer compares
* (the equivalent of a binary search), however clusterLanguages() is costly
* and the loss of accuracy from this technique is significant.
*
* This method may need to be 'fuzzier' in order to become more accurate.
*
* This function could be more useful if the universe of possible languages
* was very large, however in such cases some method of Bayesian inference
* might be more helpful.
*
* @see clusterLanguages()
* @access public
* @param string $str input string
* @return array language scores (only those compared)
* @throws PEAR_Error
*/
function clusteredSearch($str)
{
// input check
if (!Text_LanguageDetect_Parser::validateString($str)) {
return array();
}
// clusterLanguages() will return a cached result if possible
// so it's safe to call it every time
$result = $this->clusterLanguages();
$dendogram_start = $result['open_forks'];
$dendogram_data = $result['fork_data'];
$dendogram_alias = $result['name_map'];
$sample_obj = new Text_LanguageDetect_Parser($str, $this->_db_filename, $this->_unicode_db_filename);
$sample_obj->prepareTrigram();
$sample_obj->setPadStart(!$this->_perl_compatible);
$sample_obj->analyze();
$sample_result = $sample_obj->getTrigramRanks();
$sample_count = count($sample_result);
// input check
if ($sample_count == 0) {
return array();
}
$i = 0; // counts the number of steps
foreach ($dendogram_start as $lang) {
if (isset($dendogram_alias[$lang])) {
$lang_key = $dendogram_alias[$lang];
} else {
$lang_key = $lang;
}
$scores[$lang] = $this->_normalize_score(
$this->_distance($this->_lang_db[$lang_key], $sample_result),
$sample_count);
$i++;
}
if ($this->_perl_compatible) {
asort($scores);
} else {
arsort($scores);
}
$top_score = current($scores);
$top_key = key($scores);
// of starting forks, $top_key is the most similar to the sample
$cur_key = $top_key;
while (isset($dendogram_data[$cur_key])) {
$lang1 = $dendogram_data[$cur_key]['bestfit'];
$lang2 = $dendogram_data[$cur_key]['otherfit'];
foreach (array($lang1, $lang2) as $lang) {
if (isset($dendogram_alias[$lang])) {
$lang_key = $dendogram_alias[$lang];
} else {
$lang_key = $lang;
}
$scores[$lang] = $this->_normalize_score(
$this->_distance($this->_lang_db[$lang_key], $sample_result),
$sample_count);
//todo: does not need to do same comparison again
}
$i++;
if ($scores[$lang1] > $scores[$lang2]) {
$cur_key = $lang1;
$loser_key = $lang2;
} else {
$cur_key = $lang2;
$loser_key = $lang1;
}
$diff = $scores[$cur_key] - $scores[$loser_key];
// $cur_key ({$dendogram_alias[$cur_key]}) wins
// over $loser_key ({$dendogram_alias[$loser_key]})
// with a difference of $diff
}
// found result in $i compares
// rather than sorting the result, preserve it so that you can see
// which paths the algorithm decided to take along the tree
// but sometimes the last item is only the second highest
if ( ($this->_perl_compatible && (end($scores) > prev($scores)))
|| (!$this->_perl_compatible && (end($scores) < prev($scores)))) {
$real_last_score = current($scores);
$real_last_key = key($scores);
// swaps the 2nd-to-last item for the last item
unset($scores[$real_last_key]);
$scores[$real_last_key] = $real_last_score;
}
if (!$this->_perl_compatible) {
$scores = array_reverse($scores, true);
// second param requires php > 4.0.3
}
return $scores;
}
/**
* ut8-safe strlen()
*
* Returns the numbers of characters (not bytes) in a utf8 string
*
* @static
* @access public
* @param string $str string to get the length of
* @return int number of chars
*/
function utf8strlen($str)
{
// utf8_decode() will convert unknown chars to '?', which is actually
// ideal for counting.
return strlen(utf8_decode($str));
// idea stolen from dokuwiki
}
/**
* Returns the unicode value of a utf8 char
*
* @access protected
* @param string $char a utf8 (possibly multi-byte) char
* @return int unicode value or -1 if malformatted
*/
function _utf8char2unicode($char) {
// strlen() here will actually get the binary length of a single char
switch (strlen($char)) {
// for a reference, see http://en.wikipedia.org/wiki/UTF-8
case 1:
// normal ASCII-7 byte
// 0xxxxxxx --> 0xxxxxxx
return ord($char{0});
case 2:
// 2 byte unicode
// 110zzzzx 10xxxxxx --> 00000zzz zxxxxxxx
$z = (ord($char{0}) & 0x000001F) << 6;
$x = (ord($char{1}) & 0x0000003F);
return ($z | $x);
case 3:
// 3 byte unicode
// 1110zzzz 10zxxxxx 10xxxxxx --> zzzzzxxx xxxxxxxx
$z = (ord($char{0}) & 0x0000000F) << 12;
$x1 = (ord($char{1}) & 0x0000003F) << 6;
$x2 = (ord($char{2}) & 0x0000003F);
return ($z | $x1 | $x2);
case 4:
// 4 byte unicode
// 11110zzz 10zzxxxx 10xxxxxx 10xxxxxx -->
// 000zzzzz xxxxxxxx xxxxxxxx
$z1 = (ord($char{0}) & 0x00000007) << 18;
$z2 = (ord($char{1}) & 0x0000003F) << 12;
$x1 = (ord($char{2}) & 0x0000003F) << 6;
$x2 = (ord($char{3}) & 0x0000003F);
return ($z1 | $z2 | $x1 | $x2);
default:
// error: malformatted char?
return -1;
}
}
/**
* utf8-safe fast character iterator
*
* Will get the next character starting from $counter, which will then be
* incremented. If a multi-byte char the bytes will be concatenated and
* $counter will be incremeted by the number of bytes in the char.
*
* @access private
* @param string &$str the string being iterated over
* @param int &$counter the iterator, will increment by reference
* @param bool $special_convert whether to do special conversions
* @return char the next (possibly multi-byte) char from $counter
*/
function _next_char(&$str, &$counter, $special_convert = false)
{
$char = $str{$counter++};
$ord = ord($char);
// for a description of the utf8 system see
// http://www.phpclasses.org/browse/file/5131.html
// normal ascii one byte char
if ($ord <= 127) {
// special conversions needed for this package
// (that only apply to regular ascii characters)
// lower case, and convert all non-alphanumeric characters
// other than "'" to space
if ($special_convert && $char != ' ' && $char != "'") {
if ($ord >= 65 && $ord <= 90) { // A-Z
$char = chr($ord + 32); // lower case
} elseif ($ord < 97 || $ord > 122) { // NOT a-z
$char = ' '; // convert to space
}
}
return $char;
// multi-byte chars
} elseif ($ord >> 5 == 6) { // two-byte char
$nextchar = $str{$counter++}; // get next byte
// lower-casing of non-ascii characters is still incomplete
if ($special_convert) {
// lower case latin accented characters
if ($ord == 195) {
$nextord = ord($nextchar);
$nextord_adj = $nextord + 64;
// for a reference, see
// http://www.ramsch.org/martin/uni/fmi-hp/iso8859-1.html
// À - Þ but not ×
if ( $nextord_adj >= 192
&& $nextord_adj <= 222
&& $nextord_adj != 215) {
$nextchar = chr($nextord + 32);
}
// lower case cyrillic alphabet
} elseif ($ord == 208) {
$nextord = ord($nextchar);
// if A - Pe
if ($nextord >= 144 && $nextord <= 159) {
// lower case
$nextchar = chr($nextord + 32);
// if Er - Ya
} elseif ($nextord >= 160 && $nextord <= 175) {
// lower case
$char = chr(209); // == $ord++
$nextchar = chr($nextord - 32);
}
}
}
// tag on next byte
return $char . $nextchar;
} elseif ($ord >> 4 == 14) { // three-byte char
// tag on next 2 bytes
return $char . $str{$counter++} . $str{$counter++};
} elseif ($ord >> 3 == 30) { // four-byte char
// tag on next 3 bytes
return $char . $str{$counter++} . $str{$counter++} . $str{$counter++};
} else {
// error?
}
}
}
/* vim: set expandtab tabstop=4 shiftwidth=4 softtabstop=4: */
?>
