Publicação
Data-driven adaptive chess tutor bot
| Resumo: | Nowadays, more and more people are playing chess and although bots for this game are not new, the tutor type of bot is still not very developed or accessible. This work aims to develop a chess tutor bot that is open-source and takes into account the current player’s abilities to facilitate the learning process. The general structure of the bot is as follows. Several Random Forest models were trained, each with a subset of the data that were divided by player rating. Then, using Minimax during the game, the bot uses the model most suited to the player’s rating to suggest the best move. Finally, the bot uses a Large Language Model to generate comments on the moves played by the player, using information extracted from the board of the game. To increase the models’ accuracy, some experiments were made using different parameters for training. The bot was then tested by a group of eight chess beginners, each doing a game of chess while receiving comments from the bot, with overall positive results. This work is a first step towards a more personalized tutor bot which could be useful for helping beginners learn, especially when professional human tutors are not available. |
|---|---|
| Autores principais: | Lima, Pedro Miguel Marques |
| Assunto: | Chess Tutor Bot Adaptive Data driven Machine learning Artificial intelligence Search algorithms |
| Ano: | 2024 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Aveiro |
| Idioma: | inglês |
| Origem: | RIA - Repositório Institucional da Universidade de Aveiro |
| Resumo: | Nowadays, more and more people are playing chess and although bots for this game are not new, the tutor type of bot is still not very developed or accessible. This work aims to develop a chess tutor bot that is open-source and takes into account the current player’s abilities to facilitate the learning process. The general structure of the bot is as follows. Several Random Forest models were trained, each with a subset of the data that were divided by player rating. Then, using Minimax during the game, the bot uses the model most suited to the player’s rating to suggest the best move. Finally, the bot uses a Large Language Model to generate comments on the moves played by the player, using information extracted from the board of the game. To increase the models’ accuracy, some experiments were made using different parameters for training. The bot was then tested by a group of eight chess beginners, each doing a game of chess while receiving comments from the bot, with overall positive results. This work is a first step towards a more personalized tutor bot which could be useful for helping beginners learn, especially when professional human tutors are not available. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.