“Talking” into a device is no longer a strange thing to do. Speech-to-text technology, which converts spoken language into digital text, has rapidly improved in both consumer and educational products.
As the technology becomes more widely used in the classroom, state education agencies are grappling with how to handle speech-to-text technology on assessments, including how to work with third-party vendors, address security concerns, and develop policies around its use, said Sheryl Lazarus, associate director of the National Center on Educational Outcomes.
Representatives from SEAs, LEAs, and testing organizations convened in June to discuss challenges and implications related to speech-to-text technology. From that forum, NCEO recently issued the report Forum on Speech-to-Text and Scribing: Getting a Handle on What This Means.
“We don’t have all the answers, but this was a really good session to identify what some of the issues are and what kind of research is needed,” Lazarus said.
Clarity needed on who requires speech-to-text
According to state representatives at the forum, requests for speech-to-text technology have spiked in the last few years. In Ohio, for example, requests went from zero in 2015 to 67 in 2018, said Andrew Hinkle, an education program specialist with the Ohio Department of Education, who spoke during the forum.
One challenge states and school districts face is the lack of clarity around who requires speech-to-text as an accommodation.
“Beyond students with certain physical disabilities, there are also students who for a variety of other reasons may find speech-to-text very useful,” Lazarus said.
“I’m thinking about the child who can think faster than they can type or who struggles with keyboarding skills. When is that appropriate?” she said. “We need more clarity there.”
State policy varies on permissible features
State policies on the use of speech-to-text vary widely, Lazarus said. “Some states have very, very detailed specifications in what is allowed and what is not allowed, such as which types of software to use and whether spelling and grammar checkers must be turned off,” she said.
For instance, Florida policy states that speech-to-text software must be able to function without the internet and if autocorrect spelling is available, students may use it, but macro or word prediction must be turned off. In contrast, Arizona prohibits the use of speech-to-text technology on state assessments.
Implementation often requires additional training, staffing
A major implementation issue is the additional time needed to train students on how to use speech-to-text technology. In some cases, students are using a different software on assessments than what they use in the classroom.
“One of the take-aways from the forum was trying to get more of an alignment and consistency between what’s being used for instruction and what’s being used for assessments,” Lazarus said.
Students need time to practice using the technology on the assessment. The software also needs to be trained to recognize the student’s speech and can have difficulties recognizing accented speech, which can be an issue for English learners with disabilities, she said. Also, a student using the speech-to-text often needs to be tested in a one-on-one environment, which requires additional staffing.
Test security concerns persist
For speech-to-text software to run on a testing platform, the test vendor often needs to turn off certain security constraints. That can create security and compatibility issues between the testing platform and the software.
“A lot of those basic things still need to be worked out,” Lazarus said. “Discussing these issues has been really helpful for states as they work on policies, as well as for the vendors, as they think through what kinds of tech are needed and how it needs to interface with various testing platforms.”
Jennifer Herseim covers Section 504 and education technology as it relates to special education for LRP Publications.
Copyright 2018© LRP Publications, Special Ed Connection®