When the scale of the coronavirus first became apparent in February, researchers immediately set out to analyze the disease. That resulted in a fast-growing stream of submissions to Elsevier’s editorial systems. With the global situation growing ever more urgent, that research needed to be prioritized by editors and reviewers to ensure they could be published quickly while still undergoing the robust peer-review that characterises articles published on ScienceDirect.
That’s where Software Engineering Lead Mike Harris and the technology, data science and operations teams he works alongside came in. Together, they set about building a COVID-19 Submission Notifier to automatically identify relevant submissions and flag them for priority review. Mike explained:
As this research started coming in, it became clear we needed to find a way to highlight the Covid-related material for urgent action. The solution we came up with was a tool that would capture the paper title and abstract, send them off to an artificial intelligence model we designed, and that would give an estimation of how relevant the article was to Covid. If the relevancy was high, it would be flagged to the journal editors, and they would know to prioritise that research. We scoped it out and put it into action within a fortnight.
Part of the reason they were able to move so quickly was due to the working culture in Elsevier’s technology teams, he said:
We work in a highly collaborative fashion. The team is built on the values of Agile and Extreme Programming (XP), so it’s highly cross-functional and highly collaborative. There’s lots of discussion, so you’re always learning from each other, and lots of shared ownership.
Using data science, machine learning and an agile approach
For the COVID-19 Submission Notifier, Mike’s team worked especially closely with Elsevier’s data scientists. It was the data scientists who designed and trained the machine learning algorithm, while Mike and his team mapped out the overall solution and designed the API that would connect it to the submissions system:
We were in a really good position to build it as we’d been working in this problem space already, as we build support tools for Elsevier’s editorial submission systems.
We were already streaming the data, we already had the data machine learning model, so we bolted the two together with a new API to wrap the model and a little program to send an email notification – and bingo, you’ve got a solution.
For example, the team is also responsible for providing data to Elsevier’s reviewer recommendation tool, which helps find peer reviewers with the appropriate expertise to review papers when they get submitted. They also responsible for streaming information regarding additional research data sets submitted with a new manuscript in order to support Elsevier’s Open Science initiatives.
For Mike, that sense of moving quickly and making an impact is a key part of what makes his work appealing:
You really feel like you’re working on something that makes a difference. We do a lot of work with our product colleagues, which means you’re helping someone bring to life an idea they have that serves our customers. That’s very rewarding. You get good feedback, you have clear metrics so you can see where you’ve moved the needle – and you see that work you’ve done in action, hopefully making life better for the authors and editors we work with.
comments powered by Disqus