“The funding is nice — but so are the problems and data”

When it comes to AI research, Amsterdam’s universities are widely recognized for their impact — so much so that Elsevier has recruited many of their graduates.

“Sometimes we joke that we are the postdoc supplier for Elsevier,” says Dr Frank van Harmelen opens in new tab/window, Professor in Knowledge Representation & Reasoning at Vrije Universiteit Amsterdam opens in new tab/window (VU).

Frank and his VU colleagues have a long history of working with Elsevier on technology-related projects, as do faculty at the University of Amsterdam opens in new tab/window (UvA). So in 2018, the three parties took steps to formalize the relationship with the launch of the AI-focused Discovery Lab opens in new tab/window.

The Lab was established as a 5-year project, formally starting in 2020, and will run until December 2025. It has three directors: Frank, Dr Paul Groth opens in new tab/window (UvA’s Professor of Algorithmic Data Science), and Elsevier’s Lead Architect, Dr Rinke Hoekstra opens in new tab/window. Their mission is to develop intelligent services for researchers that enhance how they find and interpret scientific literature and data and formulate their hypotheses.

Elsevier, VU and UvA, with support from the Dutch government, fund three PhD students and two postdocs, who work together with Elsevier researchers in the Lab to tackle these issues via a two-pronged approach:

• Work packages: These are five research themes linked to the Lab’s goals that run for the duration of the Lab. The aim is to advance academic research in the various topic areas and publish the results.

• Spike projects: These are specific business problems identified by Elsevier teams who go on to pitch them to the Lab. Those that are selected by the Lab directors run as projects for about 12 months with the aim of producing implementable outcomes.

Developing a knowledge pipeline

Four years after its launch, the Discovery Lab has completed five spike projects, published over 50 peer-reviewed articles, presented at 40 conferences and public events, and hosted seven workshops. It is also the proud recipient of three distinguished paper awards. In addition, one PhD student has graduated, with another two expected to follow shortly.

According to Discovery Lab coordinator Dr Anita de Waard opens in new tab/window, the spike projects are a key part of the model’s success. Anita, who is also Elsevier’s VP of Research Collaborations, explains:

The Discovery Lab brings the university groups into direct contact with Elsevier’s teams — the researchers even work from the Elsevier Amsterdam office one day a week. And they can help us solve some pressing problems. Spike projects have already led to two new product ideas and two improvements to our technology infrastructure.

In fact, Anita, who has a strong research interest in scientific publishing innovation, was responsible for pitching one of the spike projects herself. It has resulted in a prototype for a tool that identifies the content of references in scientific articles. As a runner-up in a recent Elsevier idea hackathon, the concept is now being explored as a product feature on ScienceDirect.

Access to real-world challenges and datasets

For Frank and his co-Directors, however, the collaboration is about more than delivering these measurable outcomes: The opportunities to learn and share knowledge and resources are proving equally valuable.

“A phrase I often use is, ‘Sure, the funding is nice — but so are the problems and data,” says Frank. “Elsevier gives us real problems and challenges to work on, and access to great research data. And that enables us to train young people to do good science and produce quality publications that are well cited and win awards … which is the goal of every university professor.”

Cross-pollination of ideas

The Learning and Reasoning Group opens in new tab/window Frank leads at VU is comprised of 25 researchers, many of whom are funded by research projects (two of them by the Discovery Lab). But as Frank reveals: “There are absolutely no silos — there is lots of cross talk and knowledge sharing, so everyone in the group benefits from this collaboration.”

Insight into the world of industry

Frank considers the close working relationship with Elsevier a real plus for his students’ training, saying, “It offers a unique insight into another culture.”

This has certainly proved true for Dr Romana Pernisch opens in new tab/window, a postdoc at the VU who is funded by the collaboration. “I worked on a spike project that looked at how the queries that users enter into an Elsevier health solution change over time,” she explains. “But I soon learned that tracking an individual’s use of the product wasn’t going to be simple. Hospital staff often access the solution via the same terminal and account. That’s something I wouldn’t have considered before. In academia we tend to work with perfect use cases. With Elsevier, I’ve had an opportunity to see how things work in practice.”

For Romana, whose research specialty is evolving knowledge bases, that understanding is something she can take forward as she pursues a research career in the field:

The projects I do have potential for industry, so being able to see how companies work and the problems they face is really helpful. And I’ve had a unique opportunity to collaborate directly with people at Elsevier. That’s been a great learning curve, but it’s also been good for building my network.

That learning curve works the other way too, according to Anita:

“Usually when you talk about a use case, a researcher wouldn’t know what you mean. But equally, if a researcher says I’m busy reviewing papers for a conference, someone at Elsevier might not realize the pressure that puts them under. Working together like this increases our understanding of each other’s roles and challenges.”

Building a fruitful partnership

The Discovery Lab collaboration is not the only industry partnership Frank’s team is involved in, but he considers it one of the most successful. He puts this down to several factors.

“First, we are all located in Amsterdam; being in the same room can really make a difference. And there is a good understanding of AI in Elsevier — the technical knowledge is really high.”

But Frank believes the real secret to the collaboration’s success is that “unlike some other companies, Elsevier hasn’t made the mistake of thinking that we will build commercial software for them — and that’s a real feather in the Elsevier management team’s cap. We would be really bad at that type of development — you’d get professorware or PhDware.”

“Instead, we develop theories, ideas, demos and prototypes that the product teams can use. And Elsevier understands that it could be months, or even years, before there’s a potential application for some of our work.”

Inspiring Elsevier’s product teams

One thing that has surprised Frank is that the Discovery Lab is not only boosting collaboration between the partners — it’s also helping to connect like minds within Elsevier.

Anita attributes this unexpected bonus to the unique format of the spike projects. “Every time there is a pitching session, it brings together Elsevier teams with ideas about innovation. They can inspire and learn from each other, but they can also spot opportunities; for example, one pitch was picked up by another Elsevier group.”

According to Anita, other benefits for Elsevier include:

• An opportunity to explore more risky “blue sky” lines of research

• A fresh perspective on product challenges

• Access to the universities’ extensive research networks

• Insights into the latest AI science and developments

She adds: “It’s also an opportunity to work with some really bright people — and that’s inspiring and motivational for our teams.”

Enriching human knowledge with responsible AI

For Anita, another upside of the Discovery Lab is that it promotes awareness of Elsevier’s 15+ years’ experience developing and applying responsible AI to its content. “It lets people know that Elsevier is a company that is serious about AI. One of Frank’s students said: ‘Oh yes, I know Elsevier — all my friends do internships there!’ Now, we are getting people contacting us about internships before they are even announced.”

Last year, she said, Elsevier was the top choice for internships among AI bachelor’s and master’s degree students in Amsterdam, and the team interviewed 194 students, leading to 17 placements across the company.

A blueprint for collaboration

Following the success of the Discovery Lab, Anita believes the format has great potential. “We are looking at opportunities to replicate the model in other countries where we have a big IT presence,” she said. “And it’s proving an inspiration for others.

“I was at a Harvard University workshop when one of the big media-streaming platforms said they had difficulty collaborating with academia due to the different timescales and expectations," Anita said. “They were very excited when I told them about the Discovery Lab!

“This really positions Elsevier as a leader in research collaborations around AI, both nationally and internationally.

“Plans are underway to look at a next iteration of the Discovery Lab, fully harnessing the power of AI and the experience of all of the people working at and with the Discovery Lab," she added. “Because there’s a lot more to discover!”