Should MLP detect language even when language codes are not given?
Since language detection relies on langdetect not stanza, maybe MLP should include language to output even when stanza models are not available for analysis. It would be nice to still see language statistics in the cases where analysis of each of the detected language with stanza is not needed.