Semantic Scholar Academic Graph (S2AG)

该数据集包含了来自心理学、医学和计算机科学领域的105篇精选论文，其中包含了引文及其上下文信息，总共生成了70,000个引文上下文，用于细粒度的意图分类。此外，该数据集涵盖了多种不同的引文意图，包括揭示影响的（如赞扬、批评）和不揭示影响的（其他类别），并用于评估大型语言模型在意图生成任务上的性能。该数据集的规模为105篇论文，70,000个引文上下文，其任务是对引文进行细粒度的意图生成与分类。

This dataset contains 105 curated papers selected from the fields of psychology, medicine, and computer science, along with their associated citations and contextual information. A total of 70,000 citation contexts are generated for fine-grained intent classification. Furthermore, this dataset encompasses a variety of citation intent types, including impact-revealing intents such as praise and criticism, and non-impact-revealing intents (other categories). It is employed to evaluate the performance of Large Language Models (LLMs) on intent generation tasks. This dataset consists of 105 papers and 70,000 citation contexts, with its core task being fine-grained intent generation and classification for citations.

搜集汇总

背景与挑战

背景概述

Semantic Scholar学术图谱数据集（S2AG）是一个包含全学科研究论文记录的JSON格式开放数据集，与Semantic Scholar API服务深度集成，提供论文、作者和引用等学术数据。该数据集作为Semantic Scholar开放数据平台的核心组成部分，支持科研和文本挖掘应用。