disi-unibo-nlp/feast-foodex2
收藏资源简介:
该数据集包含用于根据欧洲食品安全局(EFSA)开发的食品分类和描述系统——**FoodEx2**分类法对食品项目进行分类的标注食品描述。每个示例将自由文本食品描述(`input_text`)与其正确的FoodEx2代码(`target`)和预挖掘的硬负例(`negatives`)配对,并结构化为三个互补的子任务。数据集包括训练集(82,950行,分为任务1-3)和测试集(991行),总共有83,941行。任务包括:任务1——基础术语分类(识别主要食品类别代码)、任务2——面族分类(识别适用的面族代码)、任务3——面描述符分类(识别每个面族中的描述符代码)。数据源自欧洲委员会招标文档中的公共数据集,并用于FEAST实验的可重复性研究。
This dataset contains annotated food descriptions for classifying food items according to the FoodEx2 taxonomy, a food classification and description system developed by the European Food Safety Authority (EFSA). Each example pairs a free-text food description (`input_text`) with its correct FoodEx2 code (`target`) and pre-mined hard negatives (`negatives`), and is structured into three complementary subtasks. The dataset includes a training set (82,950 rows, split across Tasks 1–3) and a test set (991 rows), with an overall total of 83,941 rows. The three tasks are as follows: Task 1 – Base Term Classification (identifying primary food category codes), Task 2 – Pasta Family Classification (identifying applicable pasta family codes), and Task 3 – Pasta Descriptor Classification (identifying descriptor codes within each pasta family). This dataset is sourced from a public dataset within the tender documents of the European Commission, and is used for reproducibility research of the FEAST experiment.




