--On Saturday, April 09, 2016 18:46 +0000 "Fred Baker (fred)"
<f...@cisco.com> wrote:

>...
> Two comments. First, she did a pretty good job - not perfect,
> but pretty darn good. Second, just like when you see
> closed-caption on a TV, what she wrote was largely phonetic,
> and often picked a homonym or was just plain wrong. I would
> have to believe that speech-to-text software would not do
> better than she did.

Fred,

As I mentioned in the simultaneous translation thread, I've
experimented extensively with speech-to-text arrangements and
IETF audio feeds, with an without RFC-trained vocabularies.   I
haven't studies that particular human transcriber experiment,
but I've worked with a lot of very good human transcribers,
including but not limited to professional court reporters.   To
the extent to which those experiments and experience are
indicative, "not do better than she did" is vastly optimistic,
with "epic fail" being closer, especially when speakers have
"unusual" accents or odd speech pacing patterns (something of
which I'm often guilty). 

Vocabulary training does help both speech-to-text systems and
human transcribers, especially with homonym-like problems, but
can introduce their own families of "just plain wrong" symptoms.



--On Saturday, April 09, 2016 21:03 -0400 Michael Richardson
<mcr+i...@sandelman.ca> wrote:

> Brian Rosen <b...@brianrosen.net> wrote:
>> I think that means we require presentation materials to
>> be preloaded and in common formats.
> 
> I think that we need this anyway for many other reasons.

And I think we have been saying that for years and that we will
continue to have late arrival of materials or late revisions
until and unless we develop firm rules that "no presentation
materials equals no presentation" as well as its close relative
"no agenda equals WG session cancellation".

However, speaking as a frequent offender, we also need to be
very aware that such a rule involves a tradeoff.  Conversations
about a particular set of issues within a day or so of the
presentation or in-session discussion are common.  They are
actually a natural advantage of gathering people together in one
place for a f2f meeting.  If insights gained from such
discussions cannot be factored into slides and presentations, we
all lose.   I don't know quite how to balance that against
having materials preloaded some significant time in advance, but
we probably need to keep the issues in mind.



    john


_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html.
https://www.ietf.org/mailman/listinfo/vmeet

Reply via email to