nuriyev/chess-commentary

Chess Commentary Dataset

~340K chess move-commentary pairs with human-written explanations from GameKnot.com.

Dataset Structure

Field	Description	Example
fen	Board state (FEN notation)	rnbqkbnr/pppppppp/8/8/4P3/8/PPPP1PPP/RNBQKBNR b KQkq e3 0 1
move	Move played	e4
commentary	Human-written explanation	"Opening with king's pawn to control the center"

Usage

from datasets import load_dataset

dataset = load_dataset("nuriyev/chess-commentary")

for sample in dataset["train"]:
    print(sample["fen"])
    print(sample["move"])
    print(sample["commentary"])
    break

Attribution

• Original source: GameKnot.com
• Paper: Jhamtani et al. (ACL 2018)
• Preprocessed by: nofarmo/Learn-Chess-Commentary

License & Disclaimer

This dataset is provided for research and educational purposes only.

The original comments are copyrighted by their respective authors on GameKnot.com. This redistribution does not grant any additional rights beyond fair use for research.

If you are a rights holder and wish to have content removed, please open an issue.

Citation

@inproceedings{jhamtani-etal-2018-learning,
    title = "Learning to Generate Move-by-Move Commentary for Chess Games from Large-Scale Social Forum Data",
    author = "Jhamtani, Harsh  and
      Gangal, Varun  and
      Hovy, Eduard  and
      Neubig, Graham  and
      Berg-Kirkpatrick, Taylor",
    editor = "Gurevych, Iryna  and
      Miyao, Yusuke",
    booktitle = "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2018",
    address = "Melbourne, Australia",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/P18-1154/",
    doi = "10.18653/v1/P18-1154",
    pages = "1661--1671",
    abstract = "This paper examines the problem of generating natural language descriptions of chess games. We introduce a new large-scale chess commentary dataset and propose methods to generate commentary for individual moves in a chess game. The introduced dataset consists of more than 298K chess move-commentary pairs across 11K chess games. We highlight how this task poses unique research challenges in natural language generation: the data contain a large variety of styles of commentary and frequently depend on pragmatic context. We benchmark various baselines and propose an end-to-end trainable neural model which takes into account multiple pragmatic aspects of the game state that may be commented upon to describe a given chess move. Through a human study on predictions for a subset of the data which deals with direct move descriptions, we observe that outputs from our models are rated similar to ground truth commentary texts in terms of correctness and fluency."
}