[bubak]

I see, thanks. As a matter of fact, in turned out that the article contains all the text, even three times (!), but it is obfuscated by some preprocessing commands. So defining get_obfuscated_article() was enough.