Could you explain why when moving from the word model to the character model the separator characters are not considered in the count?
As an example: from a template sentence "xxxxxYxxxxxYxxxxxYxxxxx", where Y is any separator character, a 3-gram in the word model could be ("xxxxx", "xxxxx", "xxxxx"), then I thought that the associated n-gram in the character model would be ("x", "x", "x", "x", "x", "Y", "x", "x", "x", "x", "x", "Y", "x", "x", "x", "x", "x") with n=3*5+(3-1)=17, since the average word length is 5 characters; while a 15-gram in the character model would be either without the "Y"s (which I think is the case) or with the "Y"s but with less than 15 "x"s.
Since the correct answer was n=15, one of these two cases is the correct one, could You explain which one and why?
Thanks.