Improve plaintext generation for language detection

This commit is contained in:
Michael 2020-10-05 12:50:18 +00:00
parent cc42c0ba27
commit 16224a7001
2 changed files with 24 additions and 1 deletions

View File

@ -1220,6 +1220,19 @@ class BBCode
return $return; return $return;
} }
public static function removeLinks(string $bbcode)
{
$bbcode = preg_replace("/\[img\=([0-9]*)x([0-9]*)\](.*?)\[\/img\]/ism", ' ', $bbcode);
$bbcode = preg_replace("/\[img\=(.*?)\](.*?)\[\/img\]/ism", ' $1 ', $bbcode);
$bbcode = preg_replace("/\[img\](.*?)\[\/img\]/ism", ' ', $bbcode);
$bbcode = preg_replace('/([@!#])\[url\=(.*?)\](.*?)\[\/url\]/ism', '', $bbcode);
$bbcode = preg_replace("/\[url\](.*?)\[\/url\]/ism", ' ', $bbcode);
$bbcode = preg_replace("/\[url=[^\[\]]*\](.*)\[\/url\]/Usi", ' $1 ', $bbcode);
$bbcode = preg_replace("/\[url\](.*?)\[\/url\]/ism", ' ', $bbcode);
return $bbcode;
}
/** /**
* Converts a BBCode message to HTML message * Converts a BBCode message to HTML message
* *

View File

@ -2476,7 +2476,17 @@ class Item
return ''; return '';
} }
$naked_body = BBCode::toPlaintext($item['body'], false); // Convert attachments to links
$naked_body = BBCode::removeAttachment($item['body']);
// Remove links and pictures
$naked_body = BBCode::removeLinks($naked_body);
// Convert the title and the body to plain text
$naked_body = trim($item['title'] . "\n" . BBCode::toPlaintext($naked_body));
// Remove possibly remaining links
$naked_body = preg_replace(Strings::autoLinkRegEx(), '', $naked_body);
$ld = new Language(); $ld = new Language();
$languages = $ld->detect($naked_body)->limit(0, 3)->close(); $languages = $ld->detect($naked_body)->limit(0, 3)->close();