Speech Recognition

Speech recognition, another new technology supporting text entry, is becoming increasingly available, affordable, and effective. It allows users to sidestep both handwriting and keyboarding and enter text into a computer by speaking into a microphone. The computer then digitizes the voice and compares it to stored information, producing on-screen text of what the user has said. In essence, speech recognition "takes dictation" from the writer. This long-awaited capability is now available in a form that, while still imperfect, is increasingly usable and is improving rapidly. Speech recognition software such as ViaVoice, Outspoken, and Naturally Speaking are compatible with desktop computers and are now available for under two hundred dollars.

To use speech recognition, writers must first train the system to recognize their voices by speaking controlled texts into the microphone. The machine matches the reader's voice to its stored list and uses the match as the basis of subsequent recognition. The training requirement, in some cases demanding more than an hour of careful reading, is a limitation of speech recognition. Some students lack the patience or the reading skills to complete the training. Dividing training into shorter time periods and coaching students on the required reading passages can help with this. Another concern with speech recognition is the somewhat cumbersome procedure for correcting errors made by the computer. Finding and correcting these errors is another skill set that students must automatize for optimal use of the system. The most promising solution to problems with training and accuracy is continued advances in technology itself, which appear to be rapid and ongoing.

Despite the training and correction obstacles, teachers and students are beginning to use speech recognition. In addition to the obvious elimination of the need to hand write or to keyboard, this technology also provides powerful support in the spelling area: if the word is recognized accurately, the computer always selects the correct spelling. Thus, writers can represent their ideas in one medium, speech, and have the computer convert it into another, text. Early research at both elementary and post secondary levels suggests that speech recognition holds promise for students with written language disorders. (Higgins & Raskind, 1995; Wetzel, 1996).