Tagging System Could Help Computers Compare Research Results

Related materials

Article: Lost in a Sea of Science Data

Printer friendly	E-mail article	Subscribe	Order reprints
	Discuss any Chronicle article in our forums
Latest Headlines Higher-Education Price Index Rises 3.6 Percent The measure of colleges' inflation costs grew only slightly more than last year, but it doesn't include recent jumps in some expenses, like utilities. Higher-Ed Bill Slowed by Two Disputes Campus Planners Discuss How to Attain Sustainability Ruling Says Colorado Board Discriminated Against University Universities in Beijing Tighten Security for Olympics U.S. Universities Negotiate Tricky Terrain in the Middle East Newly Found Chameleon Lives Fast, Dies Young Commentary Richard H. Hersh and Richard P. Keeling: On a 'Liberal Education'

By BROCK READ

Computers are tireless number crunchers and prodigious data repositories. But when it comes to scientific research, the machines have an Achilles' heel, says Ross D. King, a professor of computer science at the University of Wales at Aberystwyth: They struggle to make sense of even the most clearly written research reports.

Climatologists, for example, might use supercomputer simulations to test a theory on the effects of global warming. But if they wanted to compare their findings with those of relevant biological or botanical experiments, chances are they would have to sift through those papers by hand. That's a waste of time, according to Mr. King, and an impediment to the sharing of scientific research.

So Mr. King and Larissa Soldatova, also a computer-science professor at Aberystwyth, have designed a framework that could help computers understand and analyze scientific papers and make useful comparisons among them. The framework — an ontology, as the researchers call it, named EXPO — allows scientists to document the essential details of their experiments with keyword descriptions known as metadata.

A software program designed by Mr. King and Ms. Soldatova can then "harvest" those metadata tags, using them to look for specific kinds of research and compare the researchers' results.

According to Mr. King and Ms. Soldatova, the metadata their program relies on describes "generic knowledge about scientific experimental design, methodology, and results representation" — information about what hypothesis was formulated, how it was tested, and what the results were.

The researchers made a point to keep the tagging data general because they want EXPO to pull together research from different sciences. Scholars in several disciplines have worked to create their own ontologies, Mr. King says, "but they're doing that sort of individually in their own little silos."

"We tried to identify the key objects that are common to all experiments, whether they're biology, chemistry, or what have you," he says.

Computers that routinely read research reports might help keep scientists abreast of the work of scholars in other fields, and they could keep researchers from duplicating experiments conducted by their colleagues, Mr. King said.

Of course, scientists must first be persuaded to start using the EXPO framework — which, for the time being, requires people who write research reports to input all of the required metadata themselves.