Closed thatbudakguy closed 2 years ago
the visualization script reveals that we're predicting a reading for punctuation, etc:
it actually looks like bjuw is being used as a catchall "low-confidence" prediction or something.
bjuw
this is happening because bjuw is the first tag added to the pipe and thus has index/label 0; when the Tagger makes an empty prediction it returns a 0 which then maps to this value.
the visualization script reveals that we're predicting a reading for punctuation, etc:
it actually looks like
bjuw
is being used as a catchall "low-confidence" prediction or something.