The judges have selected a shortlist of four for the 2024 ALPSP Award for Innovation in Publishing. This year's awards are sponsored by PA EDitorial.
The finalists will be showcased in a lightning session at the ALPSP Conference on 11 September. The winners will be announced at the ALPSP Conference Awards Dinner on 12 September in Manchester.
In this series, we hear from each of the finalists.
Tell us about your organization
We love working with data, but we prefer working with you. Clear Skies was founded to support publishers. We believe that a robust, healthy peer-review system is critical to the future of research integrity. Our services are designed to support the challenging work publishers do to maintain that system while limiting the risk presented by research fraud.
This is not a "move fast and break things" startup. We believe strongly in responsible science and ethical methods. We move with care and we respect that, behind every paper (real or fake) there are real people whose careers depend on fair consideration. For that reason, we present our findings in context and use an evidence-based approach to surface actionable information. We also avoid surfacing information that might be misleading or harmful. Much like an editorial process - the value comes as much from what we don't put out as from what we do.
What is the project/product that you submitted for the Awards?
The Papermill Alarm
Tell us a little about how it works and the team behind it
The Papermill Alarm is the name we give to the alerting system that sits on top of our analytics.
That gives us an early-warning system which is available as an API through a direct agreement, or via our partners at STM, Morressier, and others (to be announced).
Sitting underneath that simple output, we have several pipelines of advanced analytics which feed into our dashboard Web Application. The dashboard surfaces important leads for investigation. What you don't see are the 50 or so different methods for detecting problematic papers that we tested and discarded. We retain only the few methods that provide genuine predictive power. Those methods include advanced AI textual analysis and network analysis.
The system improves perpetually as new data comes in from publishers. It is self-improving and self-correcting. At this stage, the underlying data is a completely unique resource describing the papermill problem in detail across the whole industry.
The Papermill Alarm was created, built, tested, and deployed at industry-scale by Adam Day. However, around 20 individuals have contributed to Clear Skies and we have built a specialised core team over the last couple of years. We give particular thanks to Adrian Stanley whose support has been incredible.
In what ways do you think it demonstrates innovation?
It was the first of its kind. The Papermill Alarm came to market in 2022 before there was any automated service dedicated to papermill detection. It's easy to see that this kind of scalable method is possible now, but at the time it wasn't clear and it took considerable testing and iteration to get there. From that initial market of 1, it has inspired numerous others to create their own papermill-detection services.
What are your plans for the future?
The future is collaboration. We want to develop our team and work with everyone throughout the research community: first, to shine a light on the problem, but secondly to help tackle it as a team with an informed perspective.
Indeed, currently, we are developing our team to reach the next level. Stay tuned!
Image caption: a semantic space showing Papermill Alarm
alerts for a publisher. Each dot is a paper. Red and orange regions represent
areas of concern.
About the author
No comments:
Post a Comment