Implementation of Speech Recognition Software for Text Processing: An Exploratory Analysis

The rationale behind implementing new information technologies is often to gain productivity improvements associated with the substitution of machinery for labor. However, the literature shows little direct evidence of a positive relationship between information technology investment and subsequent productivity benefits. This thesis reports on the examination into the productivity implications of implementing speech recognition software in a text processing environment. More specifically, research was conducted to compare text processing speeds and error rates using speech recognition software versus the keyboard and mouse. Of interest was the time required to input and proofread text processing tasks as well as the number of errors generated using both methods of text input. The empirical data offer somewhat mixed results. While users initially entered text faster using speech recognition software (p \u3c .05), they generated more errors and consequently performed proofreading and error corrections slower using speech. These results suggest that, in terms of accurate text processing, speech recognition software is still not a practical alternative to the keyboard. Therefore, implementation of speech recognition software is unlikely to result in any gains in productivity that would serve to justify its cost