function GuessLanguage($text, $lang = false)
{
if (strlen($text) <= 0) {
return false;
}
static $cache = array();
if (empty($cache)) {
$cache[] = "en";
//English is always in mind and on the first place
$rsLanguages = CLanguage::GetList($b = "", $o = "");
while ($arLanguage = $rsLanguages->Fetch()) {
if ($arLanguage["LID"] != "en") {
$cache[] = $arLanguage["LID"];
}
}
}
if (is_array($lang)) {
$arLanguages = $lang;
} else {
$arLanguages = $cache;
}
if (count($arLanguages) < 2) {
return false;
}
$languages_from = array();
$max_len = 0;
//Give customized languages a chance to guess
foreach ($arLanguages as $lang) {
$ob = CSearchLanguage::GetLanguage($lang);
$res = $ob->PreGuessLanguage($text, $lang);
if (is_array($res)) {
return $res;
} elseif ($res === true) {
return false;
}
}
//First try to detect language which
//was used to type the phrase
foreach ($arLanguages as $lang) {
$ob = CSearchLanguage::GetLanguage($lang);
$arScanCodesTmp1 = $ob->ConvertToScancode($text, true);
$arScanCodesTmp2_cnt = count(array_filter($arScanCodesTmp1));
//It will be one with most converted chars
if ($arScanCodesTmp2_cnt > $max_len) {
$max_len = $arScanCodesTmp2_cnt;
$languages_from = array($lang => $arScanCodesTmp1);
} elseif ($arScanCodesTmp2_cnt == $max_len) {
$languages_from[$lang] = $arScanCodesTmp1;
}
}
if ($max_len < 2) {
return false;
}
if (count($languages_from) <= 0) {
return false;
}
//If more than one language is detected as input
//try to get one with best trigram info
$arDetectionFrom = array();
$i = 0;
foreach ($languages_from as $lang => $arScanCodes) {
$arDetectionFrom[$lang] = array();
$ob = CSearchLanguage::GetLanguage($lang);
$arDetectionFrom[$lang][] = $ob->HasTrigrams();
$arDetectionFrom[$lang][] = $ob->CheckTrigrams($arScanCodes);
//Calculate how far sequence of scan codes
//is from language model
//$deviation = $ob->GetDeviation($arScanCodes);
//$arDetection[$lang_from_to][] = $deviation[1];
//$arDetection[$lang_from_to][] = intval($deviation[0]*100);
//Delay till compare
$arDetectionFrom[$lang][] = $ob;
$arDetectionFrom[$lang][] = $arScanCodes;
$arDetectionFrom[$lang][] = $i;
$i++;
}
uasort($arDetectionFrom, array("CSearchLanguage", "cmp"));
//echo "<pre>";foreach($arDetectionFrom as $i=>$ar){var_dump($i); print_r(array($ar[0],$ar[1],$ar[3],$ar[4],));}echo "<pre>";
//Now try the best to detect the language
$arDetection = array();
$i = 0;
foreach ($arDetectionFrom as $lang_from => $arTemp) {
$arScanCodes = $languages_from[$lang_from];
foreach ($arLanguages as $lang) {
$lang_from_to = $lang_from . "=>" . $lang;
$arDetection[$lang_from_to] = array();
$ob = CSearchLanguage::GetLanguage($lang);
$arDetection[$lang_from_to][] = $ob->HasBigrammInfo();
$arDetection[$lang_from_to][] = $ob->CheckTrigrams($arScanCodes);
//Calculate how far sequence of scan codes
//is from language model
//$deviation = $ob->GetDeviation($arScanCodes);
//$arDetection[$lang_from_to][] = $deviation[1];
//$arDetection[$lang_from_to][] = intval($deviation[0]*100);
//Delay till compare
$arDetection[$lang_from_to][] = $ob;
$arDetection[$lang_from_to][] = $arScanCodes;
$alt_text = CSearchLanguage::ConvertKeyboardLayout($text, $lang_from, $lang);
$arDetection[$lang_from_to][] = $alt_text !== $text;
$arDetection[$lang_from_to][] = $i;
$arDetection[$lang_from_to][] = $lang_from_to;
$i++;
}
}
uasort($arDetection, array("CSearchLanguage", "cmp"));
$language_from_to = key($arDetection);
list($language_from, $language_to) = explode("=>", $language_from_to);
//echo "<pre>";foreach($arDetection as $i=>$ar){var_dump($i); print_r(array($ar[0],$ar[1],$ar[3],$ar[4],$ar[5],));}echo "<pre>";
$alt_text = CSearchLanguage::ConvertKeyboardLayout($text, $language_from, $language_to);
if ($alt_text === $text) {
return false;
}
return array("from" => $language_from, "to" => $language_to);
}
