The Efficiency of Programming Through Automated Speech Recognition

The goal of this thesis is to present information which will support programming through speech recognition by finding a limit for the size of commands in the given language of the program. This limit is the point at which it will take too long for speech to be converted into text by the Viterbi Algorithm, based on the size of a Hidden Markov Model. The paper begins by introducing how speech recognition works, presenting some common issues when attempting to program by voice, and the overall motivation behind the research presented. A history of speech recognition is shown to support how programming by voice has evolved positively over time, allowing users to do more by voice alone. The algorithms which convert speech into text are formally discussed to provide an understanding of their runtimes which will be the main focus of the experiment. The future of speech recognition is also discussed based on speculations of its advancement technologically and based on the results of the experiment itself