-
August 23, 2003
The West Australian
|
No mark for style with new system
by Susan Hewit
Forget worrying about markers who just don't understand you - they may soon be replaced by heartless computers.
Curtin University researchers claim to have invented the first computer marking system which can judge a whole class of papers based on one perfect answer.
There are no marks for relevant or sassy quotes, none for clever jokes or perfect penmanship.
The system calculates your words on whether they appear in a thesaurus, determines them against a set of nouns then breaks them all down into a mathematical formula that it compares with one answer.
It doesn't care that much about grammar and it sends shivers down the spines of English teachers.
"There are two groups of people," developer and Curtin lecturer Rob Williams said. "They think it's terrific or they are sure they don't want their essays marked by computer."
"I ask them why they are quite happy about letting a computer fly them in a plane at 30,000 feet but not so sure about letting them mark their exam."
Mr Williams' system, MarkIT, was tested on law and information technology papers. It marked the papers slightly higher than human markers but lower than an American program, which differed in that 200 papers were fed into it and collated to mark new essays.
|
-
February 18, 2004
The Australian
|
TECHNOLOGY
Computer takes on marking duties
MarkIT ready for the market
A team of researchers at Curtin University of Technology has developed a prototype computer program that automatically marks student essays.
The researchers, from Curtin Business School, are hoping to develop the program, called MarkIT, into a fully fledged model.
Robert Williams, a lecturer at Curtin's School of Information Systems, has been working on a computerised marking system for the past 12 years.
Williams says that although computerised marking is an established field in the US, it is still an emerging market in Australia.
He says that during testing and development of MarkIT, essays by second-year business-law students at Curtin were graded by both a human marker and by the computer program using a single model answer.
"The resultant grades - marked by the computer and a human - were similar," Williams says.
In a second trial, the Australian program re-graded essays which had already been graded by a US program. Williams says the results were similar but the US program needed to be fed 200 human-graded model essays for comparison before it could grade other essays.
"MarkIT is a prototype at this stage, but we've managed to prove the concept works," he says.
|
