GrantRel: Grant Information Extraction via Joint Entity and Relation Extraction

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 2674–2685 August 1–6, 2021. ©2021 Association for Computational Linguistics 2674 GrantRel: Grant Information Extraction via Joint Entity and Relation Extraction Junyi Bian 1 8 , Li Huang 1 8 , Xiaodi Huang 2 , Hong Zhou 3, Shanfeng Zhu 4 5 6 7 8 1 School of Computer Science, Fudan University, Shanghai 200433, China 2 School of Computing and Mathematics, Charles Sturt University Albury, NSW 2640, Australia 3 Atypon Systems, LLC, UK 4 Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, China 5 Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence (Fudan University), Ministry of Education, Shanghai 200433, China 6 MOE Frontiers Center for Brain Science, Fudan University, Shanghai 200433, China 7 Zhangjiang Fudan International Innovation Center, Shanghai 200433, China 8 Shanghai Key Lab of Intelligent Information Processing, Fudan University, Shanghai 200433, China {zhusf, 20110240003}@fudan.edu.cn, hzhou@atypon.com Abstract As part of scientific articles, grant informa- tion refers to funder names and their cor- responding grant numbers. Extracting such funding information from articles is of signif- icant importance to both academic and fund- ing bodies. The studies on this topic face two major challenges: 1) no high-quality bench- mark datasets; and 2) difficulties in extracting complex relationships between funders and grantIDs. In this paper, we present a novel pipeline framework called GrantRel, which consists of a funding sentence classifier, as well as a joint entity and relation extractor. For this purpose, we manually label two high- quality datasets called Grant-SP and Grant-RE, respectively. In addition, our relation extrac- tion (RE) model uses both position embed- ding and context embedding in an adaptive- learning way. The experiment results have demonstrated that our model outperforms sev- eral state-of-the-art BERT-based RE baselines as higher as 6.5% of F1 scores against the PubMed Central (PMC) test set and 3.5% of that against the arXiv test set. 1 Introduction As an element of scientific articles, grant informa- tion generally includes funder names, grant num- bers, and their relations. Specifically, a funder name refers to an agency, organization, or program which provides financial support for the research. A grantID is a numerical string by which to distin- guish one grant from another. Such grant source in- formation should be automatically identified. The reasons for this are as follows: (a) The funding bodies need to track their funding statuses; (b) For This work is funded by NIH ( DK76131, HL49277 ) and a Juvenile Diabetes Research Foundation . Juvenile Diabetes Research Foundation - NIH HL49277 NIH DK76131 Funder Name Grant ID Extracted Grant Information Figure 1: In this sentence, DK76131 and HL49277 are two grant numbers of the agency funder NIH. The Juve- nile Diabetes Research Foundation is another funder. the academic, the impact of funding agencies in the scientific literature can be measured, and agencies actively supporting specific directions can be iden- tified; and (c) The literature management systems require the funding register information. Therefore, a systematic framework that is capable of automat- ically extracting grant …

Content truncated to 500 words