TASTEset

该数据集专为命名实体识别（NER）任务设计，包含了700个食谱，其中包含超过13,000个与食品产品、数量、烹饪过程等相关实体。数据集涵盖了多种实体类型，如食品、数量、单位、过程、物理性质、颜色、味道、用途和部分等，这对现有的NER模型提出了挑战。该数据集的规模为700个食谱，其中3,788个食材经过了人工标注，其任务是命名实体识别。

This dataset is specifically designed for the Named Entity Recognition (NER) task, containing 700 recipes with over 13,000 entities related to food products, quantities, cooking processes and other relevant categories. It covers a diverse range of entity types including food, quantity, unit, process, physical property, color, taste, usage and component, which poses notable challenges to existing NER models. This 700-recipe dataset has 3,788 food ingredients manually annotated, and its core task is Named Entity Recognition.