corpus_trim function

Remove sentences based on their token lengths or a pattern match