Dataset Construction for Scientific-Document Writing Support by Extracting Related Work Section and Citations from PDF Papers

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5673–5682 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5673 Dataset Construction for Scientific-Document Writing Support by Extracting Related Work Section and Citations from PDF Papers Keita Kobayashi1, Kohei Koyama1, Hiromi Narimatsu2, Yasuhiro Minami1 1The University of Electro-Communication, 2NTT Communication Science Laboratories 11-5-1 Chofugaoka, Chofu, Tokyo, 22-4 Hikaridai, Seika-cho, Soraku-gun, Kyoto {k2231042, k2131071}@edu.cc.uec.ac.jp hiromi.narimatsu.eg@hco.ntt.co.jp minami.yasuhiro@is.uec.ac.jp Abstract To augment datasets used for scientific-document writing support research, we extract texts from “Related Work” sections and citation information in PDF-formatted papers published in English. The previous dataset was constructed entirely with Tex-formatted papers, from which it is easy to extract citation information. However, since many publicly available papers in various fields are provided only in PDF format, a dataset constructed using only Tex papers has limited utility. To resolve this problem, we augment the existing dataset by extracting the titles of sections using the visual features of PDF documents and extracting the Related Work section text using the explicit title information. Since text generated from the figures and footnotes appearing in the extraction target areas is considered noise, we remove instances of such text. Moreover, we map the cited paper’s information obtained using existing tools to citation marks detected by regular expression rules, resulting in pairs of cited paper information and text of the Related Work section. By evaluating body text extraction and citation mapping in the constructed dataset, the accuracy of the proposed dataset was found to be close to that of the previous dataset. Accordingly, we demonstrated the possibility of building a significantly augmented dataset. Keywords: Scientific Document Analysis, PDF Text Analytics, PDF Information Extraction, Corpus, Bibliometrics 1. Introduction With the development of science and technology, the number of published academic papers is rapidly in- creasing1. Therefore, when reviewing their research methodology, it is difficult for researchers to survey relevant prior works and to cite them appropriately in their own papers. Despite researchers’ best efforts in these tasks, they may still unintentionally overlook cer- tain existing research or neglect certain citations where credit is due. Consequently, much work has been put into developing systems that support the writing of sci- entific papers. To reduce the burden on researchers, the fol- lowing tasks have been defined and studied for various purposes: paper summarization aimed at reducing the reading time of already searched pa- pers (Teufel and Moens, 2002; Yasunaga et al., 2019; An et al., 2021), paper recommendation aimed at reducing the search time for unfamiliar pa- pers (Bai et al., 2019), prediction of citation worthiness aimed at improving the efficiency of writing a paper (F¨arber et al., 2018; Gosangi et al., 2021), cited docu- ment allocation (F¨arber and Jatowt, 2020), and citation text generation (Xing et al., 2020; Wang et al., 2021). Despite the large amount of work carried out in as- sisting researchers with scientific paper writing, prior studies have focused only on specific tasks, which were evaluated independently using private datasets. …

Content truncated to 500 words