Sign In

Communications of the ACM

ACM TechNews

Computer Models Aim to Classify, Help Reduce Injury Accidents

View as: Print Mobile App Share:
Purdue University Associate Professor Mark Lehto

"The [computer models'] results were comparable to the human coders," says Purdue University Associate Professor Mark Lehto.

Credit: Purdue University

Researchers at Purdue University and the Liberty Mutual Research Institute for Safety are developing computer models to search through the thousands of injury reports filed in large medical or insurance claims datasets to automatically classify those reports based on specific words of phrases. The models will help identify the most important causes of injuries so action can be taken toward reducing the burden of those injuries in society. The researchers assigned codes to injury reports from workers' compensation claims using two different models based on Bayesian methods.

"The predictions were quite good," says Purdue University professor Mark Lehto. "The results were comparable to the human coders."

Lehto says injury datasets maintained by insurance companies and other organizations have largely been unused due to the expense in hiring manual coders, but the automated system could be used to obtain important information from the injury narratives and provide new insight into the potential causes and prevention of injuries. He says the models could lead to programs that automatically code reports as they are being filed.

One model, called naive, reviews individual words, while the other, called fuzzy, looks at sequences of words and phrases in narratives.

From Purdue University News
View Full Article


Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account