CHESS MOVE COMMENTATOR
Project description
The Chess Move Commentator project, undertaken as part of the Master's program in Natural Language Processing at Paris Nanterre University, aims to develop a NLP application that automatically generates commentary for chess moves. The core functionality of the application revolves around an interface where users can input their chess moves, and in return, receive insightful commentary generated by a pre-trained and fine-tuned BART model..
The project involved several stages, starting with the collection of annotated chess games from online sources. This data was then processed to create a comprehensive dataset, which includes various notations and comments associated with each move. Utilizing the BART model, the application was designed to transform FEN (Forsyth-Edwards Notation) inputs into descriptive commentary.
The backend of the project encompasses multiple scripts and models. Initially, the BART model was used in its baseline form, but it soon became clear that fine-tuning was necessary to improve comment generation. Subsequent scripts focused on refining the model to handle chess move commentary effectively. This process also included experimenting with a multi-task approach to simultaneously generate comments and predict chess moves, although performance limitations necessitated the use of alternative models like MBART.
On the frontend, the project integrates various technologies including HTML, CSS, and JavaScript, alongside tools like chessboard.js and chess.js to create an interactive chessboard. The connection between the user interface and the backend is facilitated by a FastAPI server, ensuring smooth interaction and comment generation.
Discover more about this project and click on the button below to access the GitHub Repository.
Explore More Projects
If you're interested in exploring more projects, please select another project from the dropdown menu.