Enhancing access to scholarly publications with surrogate resources

Abstract

Digital libraries containing scholarly publications are common today. They are an invaluable source of information to students, researchers, and practitioners. However, many digital libraries expose only the article metadata like title, author names, publication date, and the abstract for free; access to full-text requires access toll. Given that journal subscription charges are sometimes prohibitive, many important publications remain beyond the access of researchers, especially in developing countries. While open access publication solves this issue, the hard reality is that many research papers are not currently available for free reading or download. In this paper, we present a novel approach to alleviate this problem. We present a technique to retrieve open access surrogates of a scholarly article when the latter is unavailable freely in a digital library. Surrogates are articles semantically close to the original articles, written by the same author(s) and give valuable insights into the paper being searched for; they address the same or a very similar problem using the same or very similar techniques. Our focus on approximate matches of scholarly articles distinguishes our application from many academic search engines. We run it on a large corpus of computer science papers and compare the results with human judgment. Experimental results show that our tool can indeed identify relevant OA surrogates of access-restricted papers.

Value creation in emerging technologies through text mining: the case of blockchain

“…After the introduction of pre-trained embedding languages such as Word2vec, BERT and SciBert (see below for references and technical discussion), many studies used a larger context to identify the meaning of words in the document. Recent applications include delineation of scientific fields such as AI (Dunham, Melot, and Murdick 2020), mathematics (Greiner-Petter et al 2020) and Covid-19 related research (Kricka et al 2020;Hope et al 2020), scientific summarisation (Zerva et al 2020), identification of scientific sources (Sanyal et al 2019), and textual analysis of financial disclosures (Siano and Wysocki 2019).…”

Label informed hierarchical transformers for sequential sentence classification in scientific abstracts

“…However, these categories are not always explicitly mentioned in an abstract and therefore, there is a need for an automatic tool that can segment an abstract into discourse categories (Banerjee et al, 2020). This segmentation can also assist in various tasks like academic writing support (Huang & Chen, 2017, 2018), scientific trend analysis (Ngai et al, 2018; Prabhakaran et al, 2016), summarization (Cohan et al, 2018; Xu et al, 2020), question‐answering (Liakata et al, 2013; Verberne et al, 2007), and search for surrogates of access‐restricted papers (Santosh et al, 2018; Sanyal et al, 2019).…”