Pinterest Query-Entity Dataset

该数据集包含了一年内搜索查询日志中提取的独特查询实体对，涵盖了包括保存和长点击在内的多种互动形式，以及与产品相关的动作，如添加购物车和结账。为了对抗流行度偏差，数据集还包括了限制同一内容配对次数的约束，确保了用户活动的健壮性表示。该数据集的规模为8万对评估样本，这些样本是从更大的数据集中抽取的。该数据集的任务是搜索查询理解和推荐。

This dataset comprises unique query-entity pairs extracted from search query logs collected over a one-year period, encompassing diverse interaction forms including saves and long clicks, as well as product-related behaviors such as adding items to shopping carts and completing checkout. To mitigate popularity bias, the dataset incorporates constraints that cap the number of pairings for identical content, thereby ensuring robust representations of user activities. This dataset includes 80,000 evaluation sample pairs sampled from a larger overall dataset. The targeted tasks of this dataset are search query understanding and recommendation.