skeegan at htctu.net
Fri May 9 17:32:35 PDT 2008
> Why not take either an electronic or scanned version of
> the script, if available, and use that to creat the captioned
> version of the podcast or video.
You are absolutely correct. If you already have the text transcript, then
there would not be the need to run an audio presentation through automatic
What we were focusing on was testing how the system worked and one of the
tests included a prepared script to assess accuracy of the system. One of
the outcomes of all our testing was that we found when people spoke
extemporaneously (or were more dynamic vocally in their presentation), then
recognition accuracy decreased. When someone read from a prepared script,
then accuracy increased. Once again, we were testing different combinations
to see "what would happen if...".
We realize that if you already have a prepared script, then you would (most
likely) not be using automated speech recognition.
More information about the athen-list