According to Vice, scientists recently used AI and machine learning to reveal new knowledge buried in old research papers.
Scientists from Lawrence Berkeley National Laboratory used the algorithm Word2Vec to comb through research papers on materials science for any information scientists may have overlooked in the past. Word2Vec was able to provide researchers with “candidates for future thermoelectric materials” during its search, which it gathered via word associations. The algorithm also did not have any training in materials science, and did not know the definition of thermoelectric either, despite its success at tracking down the new candidates.
The researchers involved with the algorithm trained it by assessing over three million abstracts related to material science and compiled a vocabulary of around a half million words. From there, researchers fed the abstracts to the algorithm, which used machine learning to “analyze relationships between words,” Vice says. “The algorithm linked words that were found close together, creating vectors of related words that helped define concepts. In some cases, words were linked to thermoelectric concepts but had never been written about as thermoelectric in any abstract they surveyed. This gap in knowledge is hard to catch with a human eye, but easy for an algorithm to spot.”
Taking Things Outside of Materials Science.
One major factor that decision makers should keep in mind when looking at case studies like this is that its applicability is virtually universal – that means, even though the Word2Vec algorithm helped researchers with materials science information, it can be used elsewhere, too. This is because the algorithm isn’t trained on a specific scientific dataset, which enables it to be applied in other verticals.
It also shows that AI and machine learning is seeping into other subjects, and is proving to be handy and used by all sorts of industry experts. Decision makers may consider keeping tabs on other use cases of AI and machine learning, especially if it pertains to their expertise and has the potential to boost it.
“You could use this for things like medical research or drug discovery,” said Vahe Tshitoyan, the lead author on the study reported by Vice. “The information is out there.”