How to train Google universal encoder that two words is not matching

Question

I\'m trying to use the Google Universal Sentence encoder with the simplest way to find similarity between two sentences as the following:

import spacy_univers