The options with ...letters work at character level, they check if he sequence of characters appears anywhere in the n-gram including inside the %[token|tokens]%. The options with …words work at %[token]% level, they check if the whole word appears between spaces in the n-gram.
all
no additional filtering will be applied
starting with letters
type a sequence of letters which each n-gram should have at the beginning, spaces can be included
ending with letters
type a sequence of letters which each n-gram should have at the end, spaces can be included
containing letters
type a sequence of letters which each n-gram should have at the beginning, in the middle or at the end, spaces can be included
starting with word
type a word that should be the first word (or token) of each n-gram, typing cat will include n-grams with cat but not with catalogue
containing word
type a word that should appear anywhere in the n-gram including the first and last postitions
ending with word
type a word that should appear as the last word of the n-gram
matching regular expressions
use regular expressions for detailed filtering, the whole n-gram is treated as a continuous sequence of characters
from this list
type the n-grams to find out how frequently they appear in the currently selected corpus, the n-gram size must match the size selected above, a space must appear between all tokens, e.g. I do n't know or he 'll go