ABSTRACT
This chapter includes a description of how speech recognition (SR) works and how it can be applied to respeaking, as well as an overview of the past, present and future of this technology. SR is different from voice recognition. Speaker-dependent SR requires users to train the software so that it learns about their pronunciation, speech patterns etc. Speaker-independent SR can recognize different speakers without any training. A language model is a probabilistic mechanism for generating text that calculates the likelihood of occurrence of a word string. Respeakers must make sure that they train the software in as similar a manner as possible to the way they intend to respeak. As far as speaker-dependent SR is concerned, the 1990s were also very prolific.
