you are right , we still have problem when the content is low on text to extract, working on it.