MMECInstruct

该数据集是一个大规模、高质量的多模态指令数据集，旨在使通用多模态基础模型（MFMs）适应电商领域。它包含了针对各种电商任务所需的视觉和文本内容。在构建此数据集时，我们遵循了多模态数据融合、任务广泛覆盖以及高质量保证的原则。涉及的电商任务包括可回答性预测、类别分类、产品关系预测、产品替代品识别、多类别产品分类、情感分析以及序列推荐等。

This dataset is a large-scale, high-quality multimodal instruction dataset designed to adapt general-purpose multimodal foundation models (MFMs) to the e-commerce domain. It contains visual and textual content required for various e-commerce tasks. When constructing this dataset, we adhered to the principles of multimodal data fusion, comprehensive task coverage, and quality assurance. The covered e-commerce tasks include answerability prediction, category classification, product relationship prediction, product substitute identification, multi-class product classification, sentiment analysis, and sequential recommendation, among others.