Currently there’s no support to exclude words in that way since sentences are considered as a whole and tokenization doesn’t necessarily occur on a word level. I recommend excluding sentences that contain the words that you’d like to remain constant. We can add an option like that on our end if that would help.