On Mon, Oct 22, 2012 at 3:05 PM, Lefteris Zafiris <span dir="ltr"><<a href="mailto:zaf.000@gmail.com" target="_blank">zaf.000@gmail.com</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div class="im">If you are able to find a reliable way of chopping speech samples in segments no bigger</div>
than 20 seconds based on silence detection, so words wont be cut in half, you might come<br>
up with something very similar to Google Voice transcription service.<br></blockquote><div><br></div><div>Unfortunately Google's transcription is vastly improved by its context comprehension (for instance, understanding that the word "phone" is likely to be followed by words like "call" or "number") and chopping up the audio, even between words, will reduce that context data for the transcriber.</div>
<div><br></div><div>Good luck, anyway. </div></div><div><br></div>-- <br>-Chris Harrington<br><div>ACSDi Office: 763.559.5800</div><div><div>Mobile Phone: 612.326.4248</div></div><div><br></div><br>