ProductNet
收藏资源简介:
ProductNet是由亚马逊创建的高质量产品数据集,旨在支持产品表示学习。该数据集包含3900个非媒体产品类别,每个类别约有40-60个产品,总计约178000个产品。数据集的创建过程采用迭代方式,结合人类标注和表示学习,利用多模态深度神经网络处理产品图像和目录信息,以提高标注速度和质量。ProductNet的应用领域广泛,主要用于产品搜索、定价和其他商业应用,旨在通过高质量的数据集提升产品分类和表示学习的准确性和效率。
ProductNet is a high-quality product dataset developed by Amazon, which is designed to support product representation learning. This dataset includes 3,900 non-media product categories, with roughly 40 to 60 products per category, amounting to approximately 178,000 products in total. The dataset is constructed through an iterative approach that integrates human annotation and representation learning, utilizing multimodal deep neural networks to process product images and catalog information, thereby improving both the efficiency and quality of annotation. ProductNet covers a wide range of application fields, and is primarily applied in product search, pricing and other commercial scenarios, with the goal of enhancing the accuracy and efficiency of product classification and representation learning via this high-quality dataset.

- 1ProductNet: a Collection of High-Quality Datasets for Product Representation Learning亚马逊 · 2019年



