Require tokenisation first? #23
Description
Right now, textstat_polarity() and textstat_valence() operate on a corpus or even character object. Should we not require tokenisation first, as we do for tokens_lookup()?
Or should we provide a default tokeniser before applying the dictionaries?