ACL-FIG

ACL-FIG是由宾夕法尼亚州立大学开发的大规模自动标注科学图表数据集，包含112,052个从约56,000篇ACL Anthology研究论文中提取的科学图表。数据集内容丰富，涵盖多种图表类型，如实验结果和分析图。创建过程涉及图表提取、聚类和自动标注三个步骤，利用深度学习技术进行图表分类。该数据集主要应用于科学图表的语义理解，支持图表检索、问答和自动标注等功能，旨在解决现有学术搜索引擎对图表信息处理不足的问题。

ACL-FIG is a large-scale automatically annotated scientific figure dataset developed by Pennsylvania State University. It comprises 112,052 scientific figures extracted from approximately 56,000 research papers in the ACL Anthology. The dataset covers a wide range of chart types, including experimental result figures and analytical diagrams. Its creation involves three key steps: figure extraction, clustering, and automatic annotation, with deep learning technologies employed for figure classification. This dataset is primarily applied to the semantic understanding of scientific figures, supporting functionalities such as figure retrieval, question answering, and automatic annotation, and aims to address the insufficient processing of figure-related information by existing academic search engines.

搜集汇总

背景与挑战

背景概述

ACL-FIG是由宾夕法尼亚州立大学开发的大规模自动标注科学图表数据集，包含112,052个从约56,000篇ACL Anthology研究论文中提取的图表，涵盖多种类型如实验结果和分析图。它通过图表提取、聚类和自动标注步骤创建，利用深度学习技术进行分类，主要应用于科学图表的语义理解，支持检索、问答和自动标注等功能，以解决学术搜索引擎对图表信息处理不足的问题。