FrontPage SiteMap RecentChanges RecentNearChanges HowTo Blog RSS

SpeechToText

AKA: SpeechRecognition.

Someday (when?) computers will automatically generate transcripts of audio conversation, recorded or in real-time. This will allow us to tele/videoconference while making it easier to generate textual summaries. Textual summaries are better than just the audio or video recording itself because they take less time to read.

SpeechToText could also be helpful for quickly reworking in an AudioWiki.

Accuracy & Use

95% accuracy someone who was there can summarize or correct?
99% accuracy readable text transcripts
100% accuracy beyond human-level speech-to-text capabilities

Problems & Approaches

BayleShanks heard a few years before 2005 that you can increase accuracy by making the problem simpler by one of three approaches:

CommunityWiki's Questions


See also TextToSpeech?, VoiceRecognition (recognizing individual's voice,) WikiCitiesFuture:Speech_recognition

CategoryMultimedia?

EditNearLinks: MeatballWiki VoiceRecognition PatrickAnderson

Languages: