How would you model the mental hops that lead from one word to the next? And how about when instead of a word, the starting point are concepts grounded explicitly or implicitly in an image? These questions, and more, were the topic of my latest research project. Working to automatically generate image-term pairs for an image-grounded, collaborative Wordle game, I looked for combinations that spark the desired type of dialogue - illuminating the participants' decision-making. The project fits the broader efforts toward natural language explainability that Prof. Schlangen’s research group at the University of Potsdam is undertaking. We will look at the method I developed from an engineering perspective, going over all the NLP concepts composing it, and touch upon a bit of linguistics theory too. Level: Beginner to the domain (already familiar with Python)