Volume's been normalised using mp3gain. I manually screen all the snippets for English, music or other noise, and obvious giveaways. Looks like a Ukraine case slipped through -- it's in my todo list to remove. Thanks!

The Portuguese one also had a word sounding like "portuguese" in it that made it trivial for me.

This is against my annotation guidelines, so it's a sneaky case that slipped through. I'll make a note to review the Portuguese samples -- thanks!

And there's a Hungarian one in which I'm pretty sure I heard "Magyar" (though I could be wrong -- I don't know any Hungarian).

The Kurdish one says Turkey which helped me guess, not sure if that is being too picky though :)

The Slovak sample is very obvious, since it says Slovak at one point.

Thanks for reporting it -- that's a bug.