[4I1-OS-26a] ゲームを題材とした高度コミュニケーション(1/2)

[4I1-OS-26a-02] Classification of Hint Acts toward Implementing Codenames into Game AI

〇Kazuya Tsubokura¹, Takashi Kuboya¹, Akihisa Sanae¹, Reon Ohashi¹, Yuma Horikawa¹, Kunikazu Kobayashi¹ (1. Aichi Prefectural University)

Keywords:Game AI, Codenames

The codenames is an incomplete information communication game in which field operatives find their own team's cards among 25 cards with words written on them based on a spymaster's hints. The spymaster needs to communicate their intent to the field operatives using only one word. Therefore, by analyzing the codenames, it is thought that not only is it useful for AI implementation, but also it is possible to model the human thought during the codenames. Therefore, in this study, the game log of the codenames was collected, and the hint acts were classified from the collected log. The result showed that 75\% of the hints were nouns and that nouns were often used as hints. This result indicates the importance of the part of speech of the hint in the card selection for the future implementation to the game AI.