This is Douglas Napoleone's Typepad Profile.
Join Typepad and start following Douglas Napoleone's activity
Join Now!
Already a member? Sign In
Douglas Napoleone
Recent Activity
A very good summary, but there are some things which should be pointed out. 1. The graph is from the Sphinx Open Source recognizer from CMU. 2. It is from 2003 3. Sphinx today is 10 years behind IBM, Nuance, Google, and MultiModal (and 6 years behind Microsoft). 4. You are proposing many strawmen and burning them. What about the places where speech works? The truth is, depending on the domain, many applications have crossed the 4% WER barrier. You just don't notice anymore. Did you know that all live closed captioning is dictated? Captioners 'respeak' what is happening and use macros. The majority of medical transcription is done with computers, and then corrected by correctionists (mainly formatting and billing code extraction/confirmation.) Ever call 411? State and City please? (that is speech recognition presenting a limited phone book to the operator). Google Grand Central has huge raving. If you still think Speech Recognition does not work, download the Dragon Dictation or Search apps for the iPhone/iPod/iPad. They are free. Try them out with some full hard sentences. Don't just try single words, or things you would not normally say. Try it for real. then decide.
Douglas Napoleone is now following The Typepad Team
Jun 21, 2010