Raising and Educating a Deaf Child

International experts answer your questions about the choices, controversies, and decisions faced by the parents and educators of deaf and hard-of-hearing children.

Question from K.O., Michigan

I am a teacher of a ten year old boy with bilateral BAHAS. In addition to an FM system used in a mainstream classroom, I want to add speech to text where the speaker’s words in real time are transcribed to text on a monitor for the student using a computer or iPad. This student reads and this additional visual reinforcement can serve as amplified feedback. The student could also edit the text to created notes. What technology exists to make this happen? I am thinking a cordless microphone for the speaker, Dragon NaturallySpeaking, and then either a computer or iPad. I also want a word prediction program, but it would be great if this was software or an app that integrated word processing with the voice to text and word prediction all in one platform. Any ideas or suggestions as to best technology to make this a reality?

Question from K.O., Michigan. Posted March 17, 2015.
Response from Michael Stinson - NTID

Speech recognition has been advanced as an accommodation option for deaf or hard-of-hearing (DHH) students. With speech recognition, spoken words or phrases are automatically converted into text. Today’s speech recognition software handles continuous speech and allows the speaker to speak naturally at a conversational pace. These features have generated interest in the use of speech recognition by schools to meet the accommodation needs of their DHH students. However, a steno-based (CART) or typing-based (e.g. C-Print) speech-to-text service is a more practical and effective support option than automatic speech recognition because the speech-to-text service produces a more accurate display of information and it formats the information in a more comprehensible manner. Although speech recognition can function effectively in a variety of situations, such as when a person dictates a memo, it functions with limited effectiveness in supporting a DHH student in a class with hearing students. In this situation, highly accurate text must be produced instantly in real-time in order for students to understand the class material.

If speech recognition is to be used as a support for deaf students in a classroom, a teacher may wear a microphone and dictate directly into the speech recognition software but there is limited accuracy, typically 85%. This level of accuracy means that, on average, more than one out of every 10 words is not correct. The limited accuracy of this use of speech recognition means that, for the text produced with speech recognition, students must devote constant effort to (a) figuring out which words are incorrect, and (b) generating the correct replacements for these incorrect words. This effort is attention that the student should be devoting to the learning of class content. One reason for limited accuracy is that, for speech recognition to correctly translate a word, the word must be in the speech recognition software’s dictionary. Thus, before each class the teacher must ensure that all words, including technical words, are in the dictionary and that the software is able to recognize the teacher’s pronunciation of the words.

Demonstrations by a company promoting a product may show higher accuracy, but efforts to apply speech recognition to regular use in the classroom have not yet been very successful.