Step through synonyms and reach today's destination word.
You have six moves.
During my senior year of high school (and in the thick of college apps season), I undertook a semester of independent study into machine learning algorithms. Exploring neural networks and natural language processing ultimately led me to imagine a game about word associations, which I quickly and crudely put together with Python and the MatPlotLib library. One year later, “Neighbors” is my attempt to develop that game into a more complete product—one that is more charming, more fun, and accessible to anyone.
Based on Stanford’s GloVe model with pre-trained word embeddings, “Neighbors” relies on the same framework that powers LLM’s and generative AI. Words are encoded into vectors, whose distances from each other in multidimensional space represent the words’ semantic and contextual similarity. Thus, the nearest vectors to a given word are its neighbors, with meanings closely tied to that of the original word in context.
I want to note that this is not the only word associations game out there; in fact, I took healthy doses of inspiration from crosswords, NYT games like Connections and Spelling Bee, and more. Still, I hope that my synonym-filled honeycomb is a somewhat worthy addition to the word game canon. Enjoy!