Can I use fast-autocomplete on web corpus where data is large and there is whole sentences instead of a single word?