Remodel events to extract word values and annotate these with part of speech and frequency information