Studeni/amazon-esci-data

Amazon Shopping Queries Dataset是一个用于改进产品搜索、排名和推荐系统的综合数据集。该数据集包含查询-产品对，并使用ESCI系统进行相关性标注。ESCI系统将产品分为四类：完全匹配（Exact）、替代产品（Substitute）、互补产品（Complement）和不相关产品（Irrelevant）。数据集分为三个配置：products、queries和sources，每个配置都有训练和测试集。数据集支持多语言（英语、日语、西班牙语），并包含丰富的产品元数据，如产品标题、描述、品牌信息和颜色信息。数据集可用于产品排名、相关性分类、替代产品检测和语义搜索等应用场景。

The Amazon Shopping Queries Dataset is a comprehensive dataset designed to improve product search, ranking, and recommendation systems. It contains query-product pairs labeled using the ESCI system, which categorizes products into four types: Exact match, Substitute product, Complement product, and Irrelevant result. The dataset is divided into three configurations: products, queries, and sources, each with its own train and test splits. It supports multiple languages (English, Japanese, Spanish) and includes rich product metadata such as product title, description, brand information, and color information. The dataset can be used for various applications including product ranking, relevance classification, substitute detection, and semantic search.