This is a Codenames bot made by William Huang, Jonathan Sham, Michael Li, and Franklin Chian. Fueled by our shared love for strategy and word games, we decided to implement a robot that could play Codenames while also learning how to improve its word choices overtime. It utilizes pre-trained ML word embedding models provided by Word2vec to predict relationships between words on the Codenames board. It has the ability to either give hints to the two users playing the game, or guess words from the Codenames board after being given hints from the user. In the future, we aim to develop our own machine learning model or fine tune the existing model on more unknown words to improve the robot's overall accuracy to create relationships between words.
Taken from the Codenames website: Codenames is a social word game with a simple premise and challenging game play. Two rival spymasters know the secret identities of 25 agents. Their teammates know the agents only by their codenames. The teams compete to see who can make contact with all of their agents first. Spymasters give one-word clues that can point to multiple words on the table. Their teammates try to guess words of their color while avoiding those that belong to the opposing team. And everyone wants to avoid the assassin.
Log in or sign up for Devpost to join the conversation.