ATD-VSO

该数据集是一个针对图结构轨迹提取的基准数据集，它由带有地理实体信息的日本旅行日志构建而成，涵盖了提及内容及其指代关系。数据集包含了专有名词、普通名词短语以及指代各类地点的指示性表达，并支持对多个任务的基线系统进行评估。该数据集规模为100篇文档，按照7:1:2的比例分为训练集、开发集和测试集。涉及的任务包括访问状态预测（Vsp）、包含关系预测（Irp）以及转换关系预测（Trp）。

This dataset is a benchmark for graph-structured trajectory extraction. It is constructed from Japanese travel logs containing geographic entity information, covering mentioned content and their coreferential relationships. The dataset includes proper nouns, common noun phrases, and deictic expressions referring to various types of locations, and supports the evaluation of baseline systems for multiple tasks. It comprises 100 documents in total, which are divided into training, development, and test sets at a ratio of 7:1:2. The covered tasks include Visit Status Prediction (Vsp), Inclusion Relation Prediction (Irp), and Transition Relation Prediction (Trp).