【24h】

Citation Role Labeling via Local, Pairwise, and Global Features

机译:通过本地,成对和全局功能的引用角色标签

获取原文
获取原文并翻译 | 示例

摘要

The citation relationship between scientific publications hasrnbeen successfully used for bibliometrics, informationrnretrieval and data mining tasks, and citation-basedrnrecommendation algorithms are well documented. Whilernprevious studies investigated citation relationships fromrnvarious viewpoints, most of them share the samernassumption that, if paper1 cites paper2 (or author1 citesrnauthor2), they are connected, regardless of citationrnimportance, sentiment, reason, topic, or motivation.rnHowever, this assumption is oversimplified. In this study,rnwe propose a novel method to automatically label thernmassive citations in the scientific repository with differentrnroles, a.k.a. citation role labeling. Unlike earlier studies, wernemploy pairwise features (similarity between citing andrncited paper) and global features (citing and cited paperrnproximity on the heterogeneous graph), in addition to localrnfeatures (information extracted solely from the citing paper,rne.g. citation textual context). Evaluation result showsrnpairwise and global features, if properly used, can be veryrnhelpful to enhance the citation role labeling performance,rnespecially when full-text data is not readily available.
机译:科学出版物之间的引用关系已成功用于文献计量学,信息检索和数据挖掘任务,并且基于文献的推荐算法已得到充分证明。虽然先前的研究从不同的角度研究了引文关系,但大多数人都具有相同的假设,即如果paper1引用paper2(或author1引用author2),则无论引用,重要性,情感,原因,主题或动机如何,它们都是相互联系的。在这项研究中,我们提出了一种新颖的方法,可以用不同的角色自动标记科学知识库中的大规模引用,也就是引用角色标签。与早期的研究不同,除了局部特征(仅从引文中提取的信息,引用文献语境)之外,还使用成对特征(被引论文和被引论文之间的相似性)和全局特征(异质图上被引论文和被引论文的邻近性)。评估结果表明,如果正确使用成对和全局特征,则对于增强引用角色标签的性能非常有帮助,特别是在全文数据不可用的情况下。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号