TORQUE
收藏arXiv2020-10-06 更新2024-06-21 收录
下载链接：
https://allennlp.org/torque.html
下载链接
链接失效反馈官方服务：
资源简介：
TORQUE是一个针对英语阅读理解的新基准数据集，专注于时间顺序问题的测试。该数据集由艾伦人工智能研究所创建，包含3.2k新闻片段和21k由人工生成的时间关系查询问题。数据集的创建过程涉及培训众包工作者来标注文本中的事件，并编写和回答查询这些事件之间时间关系的问题。TORQUE的应用领域主要集中在提高机器阅读理解系统处理时间关系问题的能力，旨在解决现有系统在处理此类问题时的不足。
TORQUE is a novel benchmark dataset for English machine reading comprehension, focusing on testing questions related to temporal order. Developed by the Allen Institute for AI, this dataset contains 3.2k news segments and 21k manually generated temporal relation query questions. The dataset creation process involved training crowdworkers to annotate events in the text, as well as compose and answer questions that query the temporal relationships between these events. The primary application scope of TORQUE is to improve the capability of machine reading comprehension systems in handling temporal relation questions, aiming to resolve the shortcomings of existing systems when processing such queries.
提供机构：
艾伦人工智能研究所
创建时间：
2020-05-01
