GameWikiSum
收藏资源简介:
GameWikiSum是一个专为多文档摘要任务设计的大型数据集,由洛桑联邦理工学院人工智能实验室创建。该数据集包含14,652个样本,主要来源于专业视频游戏评论和维基百科的游戏玩法部分。数据集的创建过程涉及从Metacritic网站爬取专业评论,并通过一系列启发式规则匹配相应的维基百科页面。GameWikiSum的应用领域主要集中在视频游戏领域的多文档摘要,旨在通过整合多个来源的信息,生成关于游戏玩法的综合描述,以支持更复杂的模型训练和摘要生成技术的发展。
GameWikiSum is a large-scale dataset specifically designed for the multi-document summarization task, developed by the Artificial Intelligence Laboratory of École Polytechnique Fédérale de Lausanne. It comprises 14,652 samples primarily sourced from professional video game reviews and gameplay sections of Wikipedia. The dataset construction entails scraping professional reviews from the Metacritic website and matching corresponding Wikipedia pages via a set of heuristic rules. The primary application of GameWikiSum lies in multi-document summarization within the video game domain, where it aims to generate comprehensive gameplay descriptions by integrating information from diverse sources to support the training of more advanced models and the advancement of summarization generation technologies.

- 1GameWikiSum: a Novel Large Multi-Document Summarization Dataset洛桑联邦理工学院人工智能实验室 · 2020年



