[ilyausorov]

Indeed, although the inference output of the model is based on the ratings input that we trained it on. And that rating input was done by American English native speakers, so this iteration of the model is centered towards those accents more than e.g. UK or Australian or other accents of English from outside the US.