Why today's tech jobs need creative minds
August 30, 2017 | 9 min read
By Alison Bert, DMA
Three Elsevier technology leaders reveal the creative side of their work with natural language processing and machine learning
Caption: Tech colleagues in Elsevier’s Philadelphia office exchange ideas often during the course of their day. Sharing ideas and expertise among people with diverse talents is key to succeeding in data science and product development. (Photo by Alison Bert)
Editor’s note: This month, Elsevier Connect is exploring “the creative face of science and medicine(opens in new tab/window).” In learning how to play chess, we learn how the pieces move and the relative value of knights and rooks and pawns. But as we progress, the creative elements emerge. We discover that we can choose an opening that will lead to a slow, cautious game with the strategic maneuvering of pieces – or a wide open board where pieces are exchanged in rapid succession. We realize that recognizing patterns is as important as cold calculation. And we learn that the lowly pawn that’s worth just one point in the beginner’s manual can win a game if strategically placed.
In any art that’s rooted in science – whether it’s chess, music or cooking – we often begin by focusing on the technical elements. But ultimately technique becomes intertwined with creativity, and that creativity can take many forms.
That’s the way it is with data science, which — if you talk with the people who practice it at the highest levels — is also an art requiring creativity.
For the data scientists at Elsevier, who build features for platforms used by researchers and clinicians around the world, machine learning and NLP (Natural Language Processing) are anything but purely technical. And while some of these technologists might chuckle at the suggestion that they are artists, it’s not long before they are using terms like creativity, brainstorming and beauty when discussing their work.
“People might think machine learning is a very technical thing, but they don’t realize the nuance, and all the choices you have,” said Dr. Marius Doornenbal(opens in new tab/window), Chief NLP Scientist at Elsevier.
Check out Elsevier Careers
Welcome to our daily jam sessions
Marius’s team uses machine learning and NLP to make Elsevier’s content “actionable” so researchers and clinicians can get answers to their questions rather than having to pour through whole articles. They do this by optimizing search and discovery while isolating key facts that are relevant for the user. For example, chemistry researchers use the Reaxys database to search for specific information in the chemistry literature, patent filings, compound properties and experimental procedures. Elsevier’s data scientists have continued to make this search faster and more productive; that means telling the computer which terms and features to seek when indexing articles, devising an algorithm the machine can learn by, and "rewarding" the machine for extracting the right information, Marius explained.
To the computer scientist, the choice of algorithms can make a big difference, but if someone makes a development – a real step forward in the design of an algorithm – that is a thing of great beauty … Those are the things that really push the boundaries.
On his team, pushing the boundaries involves combining the ideas and expertise of people with diverse talents. In fact, Marius uses a term from jazz when describing how they work together.
What I really love is when we have our daily jam sessions, and someone says, ‘OK, so I’m working on this and this today.’ And someone else jumps on it and says, ‘Well, what kind of features are you using? Have you tried a combination of adjectives and nouns in phrases? Or someone says, ‘I read this article, and they say that just looking at character sequences is enough to do the job.’ This brainstorm phase is always creative; everyone gets to come up with ideas that relate to the problem, and then explain why.
People on his team bring different perspectives to the table. Marius began his studies in linguistics, researching the intricacies of language, even getting a PhD in descriptive linguistics, while all the time applying his expertise in an applied, computational setting. It’s important to know how language works, and how to apply that understanding in a way the computer understands. Marius explained: “To learn from the data, you have to tell the machine what to look for.”
What innovation looks like in data science
Dr. Michelle Gregory(opens in new tab/window), VP of Content and Innovation at Elsevier (C&I), has a joint PhD in computational linguistics and cognitive science. Only half of the people in C&I are data scientists, however; the other half are domain experts – doctors, nurses, chemistry researchers. While they have a good understanding of the capabilities of technology, they are there to make sure the technology being developed is relevant to people in their fields. Seeing your work through the eyes of end users is crucial to innovation, she said.
What’s tricky is to know what innovation looks like in data science – and how to make innovation successful. It’s not just a matter of furthering the science and algorithm; it’s a matter of really demonstrating the impact that has on the end user.
That part comes from working closely with Elsevier’s product teams, translating what their customers want into a language data scientists – and computers – can understand. “We help them figure out what data structures and algorithms and processes can help them with certain features,” she said. “So if you have this kind of data algorithm, your customers will be able to search in this way.”
For Reaxys, that can mean enabling users to find actual chemical structures instead of text terms. For ClinicalKey(opens in new tab/window), it means enabling medical professions to find reliable information quickly as they diagnose and treat patients.
What does it take to succeed?
Most people come into data science with a background in math, science, statistics and computer science. “If you are interested in math and mathematical modeling,” Michelle said, “working in data science really gives you a place to apply big data techniques.”
But some people start out as domain experts – in medicine, life sciences or chemistry, for example. “Domain experts may encounter a problem and realize that these techniques can really help,” Michelle explained.
That domain expertise can be a real plus at Elsevier, where our science and medical content are at the core of our information analytics business.
Beyond expertise, what personal traits do you need to succeed as a data scientist? For our technology leaders at Elsevier, natural curiosity and problem solving skills tend to top the list. That was the case with Michelle.
But for her, problem solving goes beyond the intellectual and is not done in isolation. It’s about using the resources around you, she explained, whether those are tools or methods or other people on the team. And for that reason, she said, it always helps to balance technical expertise with social skills and the ability to work with a team. In her group’s daily brainstorming sessions, “there’s a lot of creative dialogue that goes on – a lot of trial and error to try and figure things out.”
Where does creativity come in?
When you shop Amazon and other retail platforms, you get automated recommendations on other products you might be interested in. Increasingly, you can get recommendations on research platforms as well. On Elsevier’s, the recommenders generate personalized suggestions on articles to read and researchers to collaborate with – and soon they will recommend job opportunities and funding sources. As Director of Search and Data Science, Dr. Bob Schijvenaars(opens in new tab/window) heads a team that creates these features for research platforms like ScienceDirect, Scopus and Mendeley(opens in new tab/window).
Developing these recommenders takes creative thinking throughout the process, Bob pointed out. First, you have to give the system features or “signals” to help it judge what’s relevant for each individual user. That information can come from articles the person has viewed or stored in their library. But it also takes nuance.
For example, it’s important to differentiate between a person’s core expertise and interests, Bob explained: “If you’re looking for a particular topic that is not your core expertise, you might be interested in review articles that give you a high-level understanding of a particular area; whereas if you’re very knowledgeable or an expert on a particular topic, then you’re only interested in the latest developments in that area, and you don’t care so much about review articles.”
While building recommenders, his team also decides which channels to use to get those recommendations to the user – like an email alert with personalized article suggestions, or recommendations that display alongside an article the user clicked on.
Of course, in data science, decisions are data-driven and continuously tested by studying the interactions of users. This makes the user the ultimate judge of whether an idea works.
A further principle of the recommender team, as Bob humorously puts it, is: “Don’t look stupid.” In a job interview, you can be brilliant, but it only takes one slip to undo the good impression you’ve worked so hard to create. In data science, Bob said, you may come up with an algorithm that provides stellar recommendations in 80 percent of the cases. But if the other 20 percent completely miss the mark, they would undermine users’ trust in the system: “So you have to find creative ways to tweak your algorithm and separate the wheat from the chaff.”
It’s likely not a coincidence that so many analogies have found their way into this story. A key attribute of creativity, according to researchers reporting the (opens in new tab/window), is the habit of associating problems and ideas from seemingly unrelated fields. Which brings us back to chess.
As in the art of data science, it comes down to putting forth your best effort – and then finding creative ways to deal with the unanticipated. In Marius's words:
You have to learn all your openings, you have to get them right. But then if someone deviates from all of those games that you’ve memorized and does something unexpected – then what do you do? That’s when personality kicks in.
Technology careers at Elsevier
Elsevier employs about 1,000 technologists around the world. They create tools that help clinicians know which treatments to prescribe, that guide researchers in making groundbreaking discoveries, that train the nurses and clinicians of the future. In a creative and experimental environment you, will find ways to apply new and emerging technologies for an inspirational audience that seizes on innovation. Learn more about technology careers at Elsevier.
What-s it like to work in Elsevier Technology
The creative face of science and medicine
There is a creative side to the work being done by researchers, technologists, clinicians — and those who aspire to be. With the ever-growing possibilities of technology, that creativity is being used to solve some of today’s toughest challenges. It also takes creativity to develop the tools and technologies that solve challenges for these students and professionals.
At Elsevier, we see that creativity every day in our employees and those they collaborate with in the world of science and health. Together, we create products and services that enable professionals and aspiring professionals to realize their inspirations and aspirations.
Find more stories on the creative face of science and medicine(opens in new tab/window).
For more stories about people and projects empowered by knowledge, we invite you to visit Empowering Knowledge.