Path Quality Assessment Dataset

该数据集包含了通过亚马逊土耳其机器人收集的概念网路径上的人类成对评分，其中包括真实评分和对评分质量进行控制的成对数据。该数据集不仅包含了用于验证响应的质量控制对，还评估了评分者之间的一致性，以评估收集到的数据的可靠性。规模上，该数据集大约有90,000个成对数据，来自约1,500个有效响应。该任务是基于人类对自然性的评估来预测路径质量。

This dataset contains human pairwise ratings on ConceptNet paths collected via Amazon Mechanical Turk, including both genuine ratings and pairwise data for rating quality control. This dataset not only includes quality control pairs for validating responses, but also evaluates inter-rater agreement to assess the reliability of the collected data. In terms of scale, the dataset contains approximately 90,000 pairwise entries sourced from around 1,500 valid responses. The task is to predict path quality based on human evaluations of naturalness.