by
Charles Lowe
Florida
State University
Currents in Electronic Literacy Spring 2001 (4), <http://www.cwrl.utexas.edu/currents/spr01/lowe.html>
Endnotes
2. Unlike earlier discrete speech recognition programs, continuous speech recognition software processes speech at the rate of normal speaking speeds. Understand that continuous speech recognition software “does not rely on word boundaries to represent phonemes,” and instead uses a “much larger list of possible phonemic representations” than in discrete-speech systems, adding “to that list after the users trains it” (Randall 236). Whereas vocabularies for discrete-speech systems contain “approximately 1,100 to 1,200 phonemic representations for every 1,000 words,” continuous speech recognition programs have “3,000 phonemic representations for each 1,000 words, or considerably more” (Randall 236). Having defined the difference here, further references to speech recognition will generally imply continuous speech recognition.