Scientific and academic research has undergone two key changes in recent years – first, the growth of the ‘publish or perish’ model has resulted in an explosion of new papers, writes Dr Jabe Wilson, Consulting Director – Text and Data Analytics, Elsevier.
Dr Jabe Wilson,
Estimates suggest that there are now more than 30,000 scientific journals and approximately 2.5 million new scientific articles published annually.
Second, and perhaps even more importantly has been the move to digital. Research isn’t the only field to have been affected by digitization but, given that researchers must keep up with the latest findings, the impact has been seismic. The sheer volume of work available is too much to manage, and so we’ve seen significant growth around indexing – coupled with various forms of analytics – to make the flood more manageable.
When supported by proper analytical frameworks, the digitization of research has made it far easier for researchers to find answers to the most pressing and complex questions because they are better able to filter out irrelevant information. One of the key tools required to make this a reality is Artificial Intelligence (AI), which, in conjunction with the efforts of human researchers, is leading to a hybrid model where humans and machines work in harmony.
A Matter of Semantics
Along with the rise of digitized research, is the move towards the creation of semantic data, i.e., a way of capturing data which is aligned to how people naturally search in the real world. Put another way, rather than reading entire articles, researchers are looking to find individual facts within a paper by searching for specific terms or ideas.
This shift has only been possible thanks to the maturing of automated approaches around ‘reading’ and extracting information from articles, as well as improvements in AI and machine learning. We’ve moved away from full human curation, to a rules-based automated indexing approach; now we’re starting to see more statistical approaches based on deep learning and machine reasoning – all of which is improving productivity by helping researchers find the right information much faster.
Semantic data is vital in effective R&D because it enables researchers to link facts across related papers, spanning different knowledge domains. As a result, insights which might not be obvious by reading a single paper, can be revealed when analyzing a series of related ones. It’s also important because researchers are increasingly reliant on linked facts to be effective; as such there is a need for bespoke analytics products, built on robust semantic databases to provide the necessary support.
Content from our partners
What Comes Next?
Our understanding of AI is only improving and so we’re going to see more sophisticated applications which should provide even greater benefits to researchers. For instance, advances around how facts from disparate scientific literature can be brought together to deepen existing knowledge or transfer knowledge across domains. These sorts of developments will result in broader acceptance of AI as a valid – and indeed necessary – part of R&D efforts.
Technological progress will also help to improve collaboration and communication between scientists around the world. Implementing an AI-driven hybrid research model could free up researchers’ time from hunting through papers for the necessary information and help create more global collaborative networks, providing researchers the benefits of larger, well-integrated, cross-domain networks of linked facts. In time, this would allow them to draw more inferences and identify patterns.
Taken together, the twin trends of growing interest in AI and improved collaboration are going to be crucial if we are going to overcome the innovation crisis currently affecting many scientific disciplines.
This is demonstrated by the stalling level of productivity in evidence, with research from Stanford University showing the effective number of researchers has grown by a factor of 23 since the 1930s, yet annual productivity has declined. The upshot is that new discoveries are becoming rarer and more expensive; an AI-human hybrid model is going to be a key part of any successful effort to overcome these issues.
All the building blocks for an AI-driven R&D future are already in place – digital transformation is happening at a lightning pace across a range of industries, and employees are increasingly comfortable with the idea of working alongside, rather than in opposition to, machines.
Moving forward, we will continue to see rapid progress around the use of AI to augment human researchers, helping them make meaningful progress towards solving some of the biggest challenges we face as a species – from climate change and resource scarcity to drug resistance and precision medicine.
