Character fields now have a length of 255
Empty values are not parsed anymore when searching for ngrams
NLTK grammar is kept in cache
Better language management in ngrams extraction