MATINF (Maternal and Infant Dataset)

Maternal and Infant (MATINF) Dataset 是一个联合标注的大规模数据集，用于中文母婴护理领域的分类、问答和总结。数据集中的一个条目包括四个字段：问题（Q）、描述（D）、类别（C）和答案（A）。从中国大型母婴护理 QA 网站收集了近 200 万对问答对，其中包含细粒度的人工标记类。作者进行自动和手动数据清洗并删除：（1）样本不足的类； (2) 描述字段长度小于问题字段长度的条目； (3) 任何字段长度超过 256 个字符的数据； (4) 人为发现的格式错误的数据。数据清洗后，用剩余的 107 万个条目构建 MATINF

The Maternal and Infant (MATINF) Dataset is a large-scale jointly annotated dataset designed for classification, question answering and summarization tasks in the Chinese maternal and infant care domain. Each entry in the dataset contains four fields: Question (Q), Description (D), Category (C) and Answer (A). Nearly 2 million QA pairs were collected from large-scale Chinese maternal and infant care QA websites, which come with fine-grained manually labeled categories. The authors conducted both automatic and manual data cleaning and removed the following data: (1) categories with insufficient samples; (2) entries where the length of the Description field is shorter than that of the Question field; (3) data where any single field exceeds 256 characters in length; (4) data with manually identified formatting errors. After the data cleaning process, the remaining 1.07 million entries were used to construct the MATINF Dataset.