phptextcat.class.php

<?php

/**
 * phptextcat.class.php Textcat main class
 * 
 * PHP TextCat is a PHP implementation of the text categorization (TextCat) algorithm 
 * presented in Cavnar, W. B. and J. M. Trenkle, `N-Gram-Based Text Categorization'. 
 * Used here as a language guesser.
 *
 * Please, contribute !
 *
 * @author Christophe Dri
 */
 
 
class PhpTextCat {
    const DEFAULT_DIR_LANGUAGES = 'languages/';
    const DEFAULT_EXTENSION_LANGUAGES = '.lng';

    protected $_languages = array();
    protected $_languages_extension = self::DEFAULT_EXTENSION_LANGUAGES;
    protected $_languages_dir = self::DEFAULT_DIR_LANGUAGES;

    public function setLanguageExt($ext){        $this->_languages_extension = $ext;    }
    public function setLanguageDir($dir){        $this->_languages_dir = $dir;          }
    
    public function __construct($languages = null) {
        if(func_num_args() > 1)
            $languages = func_get_args();
        if(!empty($languages))
            $this->loadLanguages($languages);
    }
    
    /**
     * Loads a set of langages using array of strings or multimples arguments
     *
     * @param mixed $languages
     */
    public function loadLanguages($languages){
        if(func_num_args() > 1)
            $languages = func_get_args();
        else if(!is_array($languages))
            $languages = array((string) $languages);
        
        foreach ($languages as $lng) 
                $this->loadLanguage($lng);
    }

    /**
     * Loads one langage defined by $lng string ( {$lng}.lng )
     *
     * @param string $lng
     */
    public function loadLanguage($lng){
        // Sanitize language name
        $lng = preg_replace('/[^a-z0-9_-]/i', '', $lng);
        $filename = $this->_languages_dir . $lng . $this->_languages_extension;
        // Load languages N-grams
        if (file_exists($filename))
            $this->_languages[$lng] = str_replace('_', ' ', file($filename, FILE_IGNORE_NEW_LINES));
        else
            error_log("php-textcat language file [{$lng}] not found");
    }

    /**
     *  Gives a score for the text passed as an argument, depending on loaded langages.
     *
     * @param string $text
     * @return array Associative array : ( lng1 => score1, lng2 => score2, ... )
     */
    public function rateText($text) {
        // Attribute Ranks for languages depending on text words.
        $text = strtolower($text);
        $count = array();
        foreach ($this->_languages as $lng => $ngrams) {
            $count[$lng] = 0;
            foreach ($ngrams as $ngram) {
                $count[$lng] += substr_count($text, $ngram);
            }
        }

        arsort($count);
        return $count;
    }

     /**
     *  Returns the best score langage (en,fr,etc.) for the text passed as an argument.
     *
     * @param string $text
     * @return string Langage name
     */
    public function guessLanguage($text) {
        $grades = array_keys($this->rateText($text));
        return reset($grades);
    }

}