MAPLE

该数据集名为Metadata-Aware Paper colLEction（简称MAPLE），是一个针对科学文献标注的基准数据集，它基于微软学术图谱构建而成。数据集包含了特定于不同科学领域的图结构，其中论文作为节点，引用关系作为边。特别地，该数据集专注于特定科学领域的子领域图。其规模体现在以论文为节点及其对应标签的图结构上。该数据集的任务是进行多标签节点分类。The dataset is named Metadata-Aware Paper colLEction (abbreviated as MAPLE), which is a benchmark dataset for scientific literature annotation built on Microsoft Academic Graph. It contains graph structures specific to different scientific disciplines, where papers act as nodes and citation relationships function as edges. Specifically, this dataset focuses on sub-field graphs of individual scientific domains. Its scale is embodied in the graph structure that takes papers as nodes along with their corresponding labels. The task of this dataset is multi-label node classification.

搜集汇总

背景与挑战

背景概述

MAPLE是一个用于科学文献标注的基准数据集，包含20个跨越19个学术领域的数据集，覆盖从艺术到医学的广泛学科，并提供了额外的3个带有MeSH标签的数据集扩展。该数据集的特点是包含丰富的元数据，如作者、期刊、参考文献和标签层次结构，支持极端多标签文本分类任务，并已用于评估多种机器学习模型的性能。