OCR++: A Robust Framework For Information Extraction from Scholarly Articles

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3390–3400, Osaka, Japan, December 11-17 2016. OCR++: A Robust Framework For Information Extraction from Scholarly Articles Mayank Singh, Barnopriyo Barua, Priyank Palod, *Manvi Garg, Sidhartha Satapathy, Samuel Bushi, Kumar Ayush, Krishna Sai Rohith, Tulasi Gamidi, Pawan Goyal and Animesh Mukherjee Department of Computer Science and Engineering *Department of Geology and Geophysics Indian Institute of Technology, Kharagpur, WB, India mayank.singh@cse.iitkgp.ernet.in Abstract This paper proposes OCR++, an open-source framework designed for a variety of information extraction tasks from scholarly articles including metadata (title, author names, affiliation and e-mail), structure (section headings and body text, table and figure headings, URLs and foot- notes) and bibliography (citation instances and references). We analyze a diverse set of scientific articles written in English to understand generic writing patterns and formulate rules to develop this hybrid framework. Extensive evaluations show that the proposed framework outperforms the existing state-of-the-art tools by a large margin in structural information extraction along with improved performance in metadata and bibliography extraction tasks, both in terms of ac- curacy (around 50% improvement) and processing time (around 52% improvement). A user experience study conducted with the help of 30 researchers reveals that the researchers found this system to be very helpful. As an additional objective, we discuss two novel use cases including automatically extracting links to public datasets from the proceedings, which would further accelerate the advancement in digital libraries. The result of the framework can be ex- ported as a whole into structured TEI-encoded documents. Our framework is accessible online at http://www.cnergres.iitkgp.ac.in/OCR++/home/. 1 Introduction Obtaining structured data from documents is necessary to support retrieval tasks (Beel et al., 2011). Various scholarly organizations and companies deploy information extraction tools in their production environments. Google scholar1, Microsoft academic search2, Researchgate3, CiteULike4 etc. provide academic search engine facilities. European publication server (EPO)5, ResearchGate and Mendeley6 use GROBID (Lopez, 2009) for header extraction and analysis. A similar utility named SVMHeaderParse is deployed by CiteSeerX7 for header extraction. Through a comprehensive literature survey, we find comparatively less research in document structure analysis than metadata and bibliography extraction from scientific documents. The main challenges lie in the inherent errors in OCR processing and diverse formatting styles adopted by different publishing venues. We believe that a key strategy to tackle this problem is to analyze research articles from different publishers to identify generic patterns and rules, specific to various information extraction tasks. We introduce OCR++, a hybrid framework to extract textual information such as (i) metadata – title, author names, affiliation and e-mail, (ii) structure – section headings and body text, table and figure headings, URLs and footnotes, and (iii) bibliography – citation instances and references from scholarly articles. 1http://scholar.google.com 2http://academic.research.microsoft.com 3https://www.researchgate.net 4http://www.citeulike.org/ 5https://data.epo.org/publication-server 6https://www.mendeley.com 7http://citeseerx.ist.psu.edu/ This work is licenced under a Creative Commons Attribution 4.0 International Licence. Licence details: http:// creativecommons.org/licenses/by/4.0/ 3390 The framework employs a variety of Conditional Random Field (CRF) models and hand-written rules specially crafted for handling various tasks. Our framework produces …

Content truncated to 500 words