Editor’s note: This month, we are exploring the theme of “data and efficiency in science and medicine.” The ScienceDirect personalized recommender can improve efficiency for researchers by basing article recommendations on user activity data.
Keeping up with scientific literature has been described as the single most important skill throughout a researcher’s career. But it’s no easy task, with 28,100 active peer-reviewed journals and 2.5 million new articles published each year, according to The STM Report.
Is this ocean of information possible to navigate? Interestingly, as information and data continue to proliferate in this age of technology, technology is also making it easier to keep track of what’s truly relevant. One way to do this is to provide researchers with personalized recommendations based on their online behavior. Just as auto-generated personalized recommendations have changed the way people find and interact with music and film, they now have the power to change the way researchers discover content and stay up to date.
“We speak regularly to researchers, and they tell us they need a solution to this issue of sorting through the sheer volume of information,” said Maria Baronski, Senior Product Manager for ScienceDirect at Elsevier. “We see that on average researchers spend as much as four hours a week searching for articles, five hours reading them – and only 50 percent of the articles they initially find are useful to their research.”
Behavior-based personalized recommendations work in a similar way to how Netflix and Spotify suggest shows and songs based on the things you (and people like you) enjoy. With ScienceDirect personalized recommendations, researchers have an alternative way to discover the content that is most relevant to them. Since August, ScienceDirect has gradually introduced the ScienceDirect personalized recommender – the only personalized recommender based on a major full-text research content platform.
“Currently, researchers’ success in finding what they need is largely reliant on the precision of their search query,” explained Baronski, who is responsible for personalization and recommendations on ScienceDirect. “A personalized recommender finds the right information and insights for their needs by automatically recommending relevant content without requiring the user to explicitly enter a search query. In short, by applying user data and our technology to the problem, researchers can find more relvant research and spend less time looking for it.”
Active ScienceDirect users, regardless of whether they have access to the content, are already receiving emails with personalized recommendations. Every day, thousands more join them by simply registering on ScienceDirect, which is free.
The “personalized recommender” builds on the “advanced recommender” that has been part of ScienceDirect for some time. Recommendations are generated by an algorithm based on the articles viewed and downloaded by signed in ScienceDirect researchers. Much of the technology was researched and developed at Elsevier’s Research Products Data Science hub in London, which applied machine learning technology to ensure that article recommendations improve with greater platform usage – just as with Netflix and Spotify.
As of launch, the system looks at the last 2 to 100 articles in a user’s reading history, prioritizing recent activity for the recommendation delivery. Dr. Kris Jack, Head of Data Science, explained how it works.
The personalized recommender’s algorithms automatically determines users’ research interests based on their use of ScienceDirect. By applying machine learning to ScienceDirect’s rich usage logs, we can model relationships between research articles. These models capture not only the obvious relationships but also reveal more subtle ones, such as discovering that two articles from different domains that use different terminology are actually related to one another. By applying these models to the machine-readable represention of user research interests, the personalized recommender can suggest new articles that may be relevant to users. This approach can help researchers make serendipitous discoveries that would have been difficult to retrieve using a traditional search engine.
Extensive testing has shown that using a personalized approach can highlight content that might not ordinarily have occurred to users. That inclusiveness is vital to scientists and scholars who are working to stay at the forefront of their fields. Meanwhile, as our customer research has shown, new directions in research are often sparked by insights that are serendipitously found. That exposure to a range of new ideas can be of crucial importance to researchers, highlighting research they may otherwise have missed.
“We see this as being at the vanguard of a more customized approach to the research workflow,” said Dr. Jack. “We are tailoring reading options directly to the needs of individual users – we really want to make a dent in that time spent looking for the relevant research, and to ensure that researchers have time for what matters most: their research.”
Protecting your privacy
When the ScienceDirect personalized recommender discovers relevant recommendations, researchers will receive up to one email a week with the best recommendations for them. ScienceDirect users are under no obligation to use the feature; although they will be automatically opted in when they register, they can unsubscribe via the email or use the on-off toggle on the manage alerts page. Read more about Elsevier’s approach to data privacy.
Recommender product team
Adam Darts , Adam Davidson, Benjamin Pettit, Christopher Miliotis , Claudio Bellei, Craig Scott, Daniel Kershaw, David de Kock, Diogo Caldeira-pires, Edward Ingold, Elinor English, Filippo Mortari, Hiroshi Fukunari, Isabel Martin Garcia, James Coward, Jelaxshan Jehanathan, Jennifer Van den Brink, Jonathan Warburton , Kris Jack , Leanne Tite, Venkateswarachary Mailaram, Maria Baronski, Mari C. Barredo, Maya Hristakeva, Monica Beltran, Phani Raavi, Richard Campbell, Richard Lyne, Ryan Connolly, Supriya Pandey, Susana Cuesta Diaz, Thijs Van Noort, Tun Shwe, Victoria Rao, Willie Doran