Speech Technology Magazine

 

ETS Adds Voice identification to Increase TOEFL Exam Security

By Leonard Klie - Posted Jul 10, 2012
Page1 of 1
Bookmark and Share

Educational Testing Service (ETS), the Princeton, N.J.–based creator of the TOEFL (Test Of English as a Foreign Language) exam to measure English language proficiency among foreign students applying for admission to English-speaking colleges and universities, recently introduced biometric voice identification to maintain fair and reliable testing.

The security measure provides an additional proven technique to add to the TOEFL program's comprehensive security system in authenticating TOEFL test-takers globally. The TOEFL test is recognized by more than 8,500 colleges, universities, and agencies in more than 130 countries. Globally, it is available at more than 4,500 testing sites in more than 165 countries. To date, more than 27 million students around the world have taken the exam.

Similar to the highly advanced speaker identification platforms used by government and law enforcement agencies, the voice authentication software uses statistical pattern matching techniques, advanced voice classification methods, and inputs from multiple systems to compare speech samples from TOEFL test-takers. The speaker identification system offers the ability to create voiceprints for detailed analysis to validate TOEFL test-takers.

The new technology is being used as part of test security investigations this year and will gradually be used on a larger scale in 2013 and beyond.

The test already has a spoken component—it contains four parts that rely on speaking, reading, writing, and listening. "We are already digitally capturing the voice during the test," explains Ray Nicosia, director of ETS's Office of Testing Integrity. "We're using the voice screening to make sure it's the same person throughout the test."

The inclusion of biometric voice identification technology is yet another tool in the TOEFL test security portfolio to ensure test integrity worldwide, Nicosia says. Other security steps include photo and handwriting recognition, instructor training, and even room arrangement during the test.

Though still in the pilot phase, Nicosia says ETS has already had some success with the voice technology in supporting ongoing investigations.

According to ETS, foreign students face stiff competition when applying to U.S. colleges, which has brought about a rise in the number of attempts to falsify TOEFL test scores. "There will always be people who try to get an unfair advantage during the test," Nicosia says. "This gives us more ammunition if we suspect that someone else is taking the test for you," Nicosia says. "It's an additional tool that makes it a more scientific process for us."

Judith Markowitz, president of J. Markowitz Consultants, a firm specializing in voice security applications, sees this as a "very good way" of using available voice technology.

"It's a different approach that saves [ETS] a lot of work. They don't have to do anything extra to get the voice data," she says.

Most deployments of voice security in education have been tied to distance learning, where students need to authenticate themselves before taking online classes or exams, Markowitz says. "With distance learning, you have to have a high level of authentication. You really need to make sure the person is who he says he is."


Page1 of 1