Real-Scenario Multimodal Retrieval Dataset from Taobao

KDD Cup 2020 Challenges for Modern E-Commerce Platform: Multimodalities Recall offers real-scenario data from the mobile Taobao, one of the largest e-commerce platforms. The dataset consists of Taobao search queries and product image features, which are organized into a query-based multimodal retrieval task.
You can rank a collection of candidate products based on their image features with a given search query in natural language form. Most of these queries are noun phrases searching for products with specific characteristics. The images of the candidate products are provided by the sellers displaying the product features. Candidate products most relevant to the query are regarded as the ground truth of the query, which are expected to be top-ranked by the participating models.

KDD杯2020现代电子商务平台挑战赛：多模态召回任务数据集源自全球顶级电商平台之一的移动淘宝，收录真实业务场景下的实测数据。该数据集涵盖淘宝搜索查询词与商品图像特征，并构建为基于查询的多模态检索任务。
参赛选手可基于给定的自然语言形式搜索查询词，结合候选商品的图像特征对候选商品集合进行排序。绝大多数查询词为带有特定属性描述的名词性短语，用于检索匹配对应特征的商品。候选商品的图像均由卖家上传，用于直观展示商品的属性特征。与查询词关联度最高的候选商品将被作为该查询的基准真值（ground truth），参赛模型需将其排在检索结果的最前列。