This is based on Oliver Guhr's work. The difference is that it is a finetuned xlm-roberta-base instead of an xlm-roberta-large and on sixteen languages instead of four: English, German, French, Spanish, Bulgarian, Italian, Polish, Dutch, Czech, Portugese, Slovak, Slovenian, Greek, Swedish, Danish, Hungarian and Romanian.

----- report -----

          precision    recall  f1-score   support

       0       0.99      0.99      0.99  98904114
       .       0.94      0.95      0.95   6331979
       ,       0.86      0.86      0.86   8390616
       ?       0.89      0.84      0.86    182114
       -       0.59      0.31      0.40    341375
       :       0.72      0.52      0.61    210519

accuracy                           0.98 114360717

macro avg 0.83 0.75 0.78 114360717 weighted avg 0.98 0.98 0.98 114360717

----- confusion matrix -----

 t/p      0     .     ,     ?     -     : 
    0   1.0   0.0   0.0   0.0   0.0   0.0 
    .   0.0   1.0   0.0   0.0   0.0   0.0 
    ,   0.1   0.0   0.9   0.0   0.0   0.0 
    ?   0.0   0.1   0.0   0.8   0.0   0.0 
    -   0.1   0.1   0.4   0.0   0.3   0.0 
    :   0.0   0.3   0.1   0.0   0.0   0.5