Text Data Mining: Unlocking the hidden potential from scholarly content.
- 1. 1 TDM: Unlocking the hidden potential from scholarly content
- 2. 2 Until recently, text mining has mostly been restricted to post-publication PDFs and has proved slow and difficult. The focus for scholarly content has often been limited to metadata and abstracts. TDM is evolving to extract a wealth of information that can support the entire scholarly community – from authors to publishers. Making sense of unstructured content
- 3. 3 Landscape
- 4. 4 6% YoY growth in manuscript submissions 42% authors post their preprint before journal submission 300% increase in the number of preprint servers since 2015 The research keeps growing Published work and preprints 6% 300% 42%
- 5. 5 Too many manuscripts. Not enough time. Submission to publication time expanding. 48 Hours First review round Submission to publication Screening 13 Weeks 400 Days
- 6. 6 XML often made available for Open Access articles, but not all publishers make XML available to TDM services (API). Rise of preprint servers and number of journals inviting article submission via these servers increases need to mine non-XML content. Most authors still submit manuscripts to publishers & preprint servers in Word or PDF. Some servers convert content into XML, but majority of platforms only allow for the preprint to be downloaded in the same format it was uploaded in. The format challenge
- 7. 7 Software used by authors Word still the preferred format Writing software used by authors submitting to bioRxiv. Source: Sever et al (2019) bioRxiv: the preprint server for biology. https://dx.doi.org/10.1101/833400
- 9. 9 Extracting structured content from any document Dixon WG, Beukenhorst AL, Yimer BB et al. 2019. doi:10.1038/s41746-019- 0180-3 Content extracted to a structured format
- 10. 10 Distilling research into headlines and key information Rosyadi S, Haryanto A. 2019. doi:10.31124/advance.9989639.v1 Distillation to unified format
- 11. 11 Opportunities
- 12. 12 Manuscript submission Manuscript screening Peer review Promotion TDM: What are the opportunities? TDM can work at any stage of the publishing process, opening up a huge number of opportunities from manuscript drafting and screening to promoting the published article.
- 13. 13 • Metadata extraction to automate population of submissions system (Title, author, affiliations, abstract, keywords). • Reduces author friction / duplication of effort. • Previous work in this area has focused on the biomedical domain, but this opportunity can apply to any domain. Automating submissions process
- 14. 14 • Data extraction for manuscript screening (key methods, results, sample size, participants, ethical compliance etc.) • Clear article context/overview for reviewers. • One-click access of cited sources & main findings. • Table extraction for analysis of statistical calculations. Speeding up peer review
- 15. 15 Surfacing cited sources & their main findings Krohn L, Ruskey JA, Rudakou U et al. 2019. doi:10.1101/19010991 Cited sources and their main findings surfaced
- 16. 16 • Extract, parse and link citations from archives dating back hundreds of years. • Large scale reference population of open citation networks (BMJ Case study) • Improve exposure/discovery of older research. Exposing more content through citation networks
- 18. 18 How publishers can help. Make XML available for all Open Access articles rather than just the final PDF for text mining. Enrich citation networks with additional content (e.g. abstract, highlights) in a machine-readable format. Make all cited sources more easily verifiable for authors and researchers. Converting articles & preprints into a universally structured format for more effective TDM. Allow authors to write articles natively in a machine-readable format. 1 2 3 4
- 19. 19 …equal rights for friendly bots! And finally…