Pinterest User Engagement Logs

该数据集包含了基于用户搜索行为及其相应信息流浏览日志生成的每日用户参与度记录。具体来说，数据集中包含了展示给用户的查询项对，以及这些项的特征和用户参与类型，如保存、长按、下载、截图、举报和隐藏等多种形式。为了评估模型的性能，我们采用了如HITS@K等评价指标。由于实际操作的可处理性，该数据集从近十亿的每日搜索中进行了下采样。该数据集的任务是对搜索系统进行预排序性能评估。

This dataset contains daily user engagement records generated based on users' search behaviors and their corresponding feed browsing logs. Specifically, the dataset includes query-item pairs presented to users, along with the features of these items and various types of user engagement, such as save, long-press, download, screenshot, report, hide and others. To evaluate model performance, we adopted evaluation metrics such as HITS@K. For practical operability, this dataset was downsampled from nearly one billion daily searches. The task of this dataset is to conduct pre-ranking performance evaluation for search systems.