Commentary Generation from Data Records of Multiplayer Strategy Esports Game
Zihan Wang, Naoki Yoshinaga
TL;DR
This work addresses generating textual commentaries from structured esports data for League of Legends. It constructs two large LoL-specific data-to-text datasets (LoL19 and LoL19-21) by pairing event-based JSON game records with YouTube subtitles and evaluates Transformer-based models, including T5 and Llama2 variants, under finetuning and in-context learning regimes. Automatic metrics (sacreBLEU, ROUGE-L, BERTScore, BARTScore) plus a human-driven strategic-depth evaluation reveal that larger pre-trained models improve quality and strategic insight, but challenges remain in aligning past events with current moments and maintaining context over long narratives. The study provides a new dataset resource and outlines future directions, such as cross-modal integration with visuals to enhance commentary generation for esports.
Abstract
Esports, a sports competition on video games, has become one of the most important sporting events. Although esports play logs have been accumulated, only a small portion of them accompany text commentaries for the audience to retrieve and understand the plays. In this study, we therefore introduce the task of generating game commentaries from esports' data records. We first build large-scale esports data-to-text datasets that pair structured data and commentaries from a popular esports game, League of Legends. We then evaluate Transformer-based models to generate game commentaries from structured data records, while examining the impact of the pre-trained language models. Evaluation results on our dataset revealed the challenges of this novel task. We will release our dataset to boost potential research in the data-to-text generation community.