Whether it’s plastic pollution, climate change or endangered species conservation, there’s a growing tsunami of scientific data out there – how do researchers wade through it to fill gaps, learn from failures and build on successes?
Faced with this conundrum when wanting to restore sea otters to wild habitats, US wildlife biologists turned to artificial intelligence to evaluate conservation outcomes from more than 4,000 studies over four decades.
Happily, as published in the journal Cell Press, results suggest that the success of species reintroductions is growing.
The researchers applied natural language processing, a type of machine learning that analyses strings of words to extract relevant information, matching what a human reader would take from it.
More specifically, in this instance, they used sentiment analysis to find words with positive or negative emotional values to gauge the overall rate of success or failure.
The resulting trends suggest increased conservation success, says senior author Kyle Van Houtan from Monterey Bay Aquarium in California.
“Over time, there’s a lot less uncertainty in the assessment of sentiment in the studies, and we see reintroduction projects become more successful… Looking at thousands of studies, it seems like we’re getting better at it, and that’s encouraging.”
The process itself would be a boon for scientists who trawl through laborious meta-analyses. It could also help streamline conservation initiatives and save costs, according to study co-author Lucas Joppa, chief environmental officer at Microsoft.
“Machine learning, and natural language processing in particular, has the ability to sift through results and shine a light on success stories that others can learn from,” he says.
To glean their answers, the researchers scanned more than a million words from the abstracts of 4,313 species reintroduction studies published from 1987 to 2016 and used “off-the-shelf” words assigned a sentiment score from movie and restaurant reviews to build a model that could give each abstract an overall rating.
“The scores gave us a trend over time,” says Van Houtan, “and we could query the results to see [whether] the sentiment was associated with studies on pandas or on California condors or coral reefs.”
To double-check their findings, they looked at the most common words assigned positive and sentiments and found these aligned with what they would use in their own work to determine successes or failures, for instance “success,” “protect,” “growth,” versus “threaten,” “loss,” “risk”.
The trends also matched known outcomes from reintroduction programs, such as that of the California condor.
Some wrinkles still need to be ironed out of the process. Although it’s been used for more than a decade in commercial settings, the application to academic literature is at a proof of concept stage.
But investing in its development would be worthwhile, the researchers write, “perhaps especially to ensure that scientific productivity meets practical progress”.
And it could be a welcome tool for scientists working on some of the planet’s most pressing challenges.
“There’s this plethora of data that’s right at our fingertips,” says Van Houtan, “but it’s this sleeping giant because it isn’t property curated or organised, which makes it challenging to analyse.
“We want to connect people with ideas, capacity, and technical solutions they might not otherwise encounter so we can bring some progress to these seemingly intractable problems.”