It did so by querying prompts similar to that of the one inputted from a mapping of prompts to answers. The main limitation of this system is that it can only respond to prompts similar to those in the mapping. I can’t lexically interpret or derive meaning from sentences — two synonyms are viewed as