Associated Press
CARY, N.C. — Software can translate foreign languages, turn spoken voice into words on a page and understand e-mail well enough to automatically type customized replies.
As programmers continue trying to mimic the human brain, the day may come when software can even read your e-mail and detect lies.
In a few months, SAS Institute Inc., the world’s largest private software company, will begin selling a package that could be adapted to compare word and grammar patterns to a writer’s previous work and reveal inconsistencies.
SAS, which already makes fraud-detection systems for banks and phone companies, says its Text Miner program will be sold to bolster business intelligence, not uncover falsehoods.
But with the right amount of tinkering, it’s not farfetched, the company believes.
"Is it here today? No. Is it theoretically possible if someone were to pursue it? Yes," said SAS spokesman Bob Chase.
The products are part of a growing inventory of so-called "text mining" software that seeks patterns hidden in vast data collections.
Revenue from sales of all types of data mining software — of which text mining is a subgenre — will grow from about $540 million this year to about $1.5 billion in 2005, according to market research firm IDC.
Text mining programs can write reports — and even recommend a course of action — by gleaning clues from e-mail, medical reports, news dispatches or consumer comments recorded by call center operators.
One unnamed U.S. oil company is using such a program to analyze employee comments on public Web sites to measure their reaction to management decisions, according to META Group consultants.
University of Louisville medical researchers are using SAS software to retrieve buckets of information on ailments and treatments found in medical literature.
The software can group articles on a particular kind of clinical research and reject others.
Another text mining product, made by SER Solutions Inc. of Herndon, Va., is touted for its ability to scour the Internet for key words and phrases on bulletin boards and Web pages. Germany’s Bayer AG’s German plans to use it to search for potential patent infringements on its plastics business.
The products try to mimic the brain’s ability to tread unfamiliar turf — learning, finding patterns, then sorting and analyzing what’s important and why.
So-called neural networks are good at making sense out of chaos, tolerating gaps in information and making many small decisions before coming to a conclusion.
By 2006, companies will use text mining software to analyze customer feedback, whether it comes from the Internet, call centers or sidewalk surveys, said an analyst Alessandro Zanasi, formerly of META Group.
Personnel divisions will use it to match resumes with hiring needs. Market research divisions will be able to better track competitors, sales trends and research extracted from huge volumes of patents, scientific articles and news reports, Zanasi said.
But what about managers who want software to determine whether an employee has been stealing and lying about it?
A few experts were skeptical.
"I won’t say this can’t be done. I would say that the odds are very high that if you’re basing it just on speech, the accuracy is going to be quite modest or close to chance," said Paul Ekman, a psychology professor at the University of California at San Francisco and author of "Telling Lies: Clues to Deceit in the Marketplace, Politics, and Marriage."
Copyright ©2002 Associated Press. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.
Talk to us
> Give us your news tips.
> Send us a letter to the editor.
> More Herald contact information.