Codenames AI Framework
Natural Language Processing Research, 2019
An implementation of an AI framework for the popular word-guessing game Codenames, exploring the challenges of word-vector relationships and semantic similarity in natural language processing.
Project Overview
This project implements an AI system for playing Codenames, focusing on both the Spymaster and Guesser roles. The framework uses word embeddings and vector space mathematics to:
- Generate semantically meaningful clues as the Spymaster
- Interpret clues and select appropriate words as the Guesser
- Handle negative associations with opponent and assassin words