Re: [agtk-devel] [gilles@harfang.homelinux.org: AG Questions]

[Haejoong L. <hae...@un...>]

Hi Gilles,

I've posted your message on agtk-devel mailing list.

Here are my answers for your questions.  I hope other members can
answer some of your questions too.

About agtk-devel sourceforge list:  Steven, I think we can make the
mailing list public, can't we?

1. I don't think we have Pratt <-> AG filter.  Because of the limit on
our resources, I wonder if we will be developing such filter in near
future.  Thus contribution from the community is always more than
welcome.  If you can write c++ filters for AGTK, that's great.  If you
can write Python or Tcl filters, that is also great, although we will
have to rewrite using c++.

2. I think the AG data model itself makes "distributed AG" possible.
Maybe Xiaoyi can answer this question regarding the source code.

3. We wrote Transcriber input/output filter for internal use.  Nicolas
Garbe took this filter to check, and it didn't work for him.  I asked
him for a precise format description, and I didn't get the answer.
So, this work is on hold.

4. Annotation type is used to make classes of annotations like
phoneme, word and sentence as in TIMIT examples.  Annotation feature
adds information about the object annotated by the annotation.  Since
there is no restriction on how the feature is used, it's okay a
feature has a field for annotation type.  There must be a special
reason for doing so in the examples you've seen.

5. As long as I concern, if it's efficient for large data and we can
implement Kleene closure operation [http://arXiv.org/abs/cs.CL/0204004]
easily, there is no problem.  I believe Steven has something to say on
this issue.


Haejoong

P.S. To agtk-devel members: Please don't forget to CC to
gi...@ha...


On Tue, Nov 19, 2002 at 02:22:51PM -0500, Haejoong Lee wrote:
> ----- Forwarded message from Gilles Sadowski <gi...@ha...> -----
> 
> Date: Tue, 19 Nov 2002 00:16:47 +0100
> From: Gilles Sadowski <gi...@ha...>
> To: hae...@un...
> Subject: AG Questions
> 
> Hi.
> I sent you the following mail 5 days ago, and only today I got a message 
> from my mail server that it hadn't been able to deliver it!  So I try 
> again in the hope you'll get it this time.
> Best regards.
> 
> Haejoong Lee wrote:
> 
> >I'm do sorry, I thought you were on the mailing list.
> >Here is Steven's answer for your question. -- Haejoong
> >
> >----- Forwarded message from Steven Bird <sb...@cs...> -----
> >
> >To: Haejoong Lee <hae...@un...>
> >cc: Kazuaki Maeda <ma...@un...>,
> >   Xiaoyi Ma <xm...@un...>, agt...@li...
> >From: Steven Bird <sb...@cs...>
> >Subject: Validating feature values
> >Date: Fri, 18 Oct 2002 11:32:03 EDT
> >
> >[...]
> >
> Thanks for the answer.  Indeed I only subscribed to 'agtk-announce'.
> Right now, I wanted to subscribe to 'agtk-devel' but, strangely, 2 of
> the 3 agtk mailing lists don't show up on the sourceforge page!
> 
> So I hope you won't mind my asking another round of questions directly
> to you.
> 
> 1.  You probably know the speech analysis software "Praat".  Are there
> some filter tools that would enable data interchange between Praat and
> AGTK tools?  That could be something like a converter from Praat 'Text
> File' format to an "ag.dtd" conformant document.  But a more elaborate
> access would be through a "standard" module that could insert Praat data
> via any AGTK-based annotation tool...
> Going the other direction (AGTK -> Praat) would be interesting too (e.g.
> to do a tone analysis on selected parts of a speech signal being
> annotated with AGTK, and directly insert the result...), using the
> 'sendpraat' utility program...
> 
> 2.  Does AGLIB have provision for dealing with "distributed" AG.  By
> distributed, I mean, for example, 2 documents referring to the same
> signal, but containing different types of <Features>?  I already asked
> this question, and I think the answer was 'yes', but could you point out
> where in the code this is dealt with?
> 
> 3.  Is there a tool for converting 'Transcriber' data into AG?
> 
> 4.  Could you explain the difference between the 'type' attribute in
> <Annotation> and <Feature> (because I've noticed that in many examples,
> the 2 are in fact assigned the same value)?
> 
> 5.  About the problem of validating <Feature>s, Steven Bird said:
> 
> > [...] An ongoing research activity is to define an AG query language 
> which can be
> > used to make declarations of well-formedness, then compile this into SQL
> > for efficient execution.  This would make it possible for any AG data 
> file
> > to document the details of its structure (by including a query expression
> > in its header, say), and likewise for AG tools to document the 
> assumptions
> > they make about AG data.  Going a step further, we might be able to
> > automatically determine what tools could be applied to what AG data.
> > Unfortunately this is all pie in the sky right now.[...]
> 
> 
> Do you think the W3C XQuery language (and, somehow, RDF) could be used 
> for that purpose?
> 
> 
> Many thanks for your time.
> 
> Gilles Sadowski
> 
> 
> 
> ----- End forwarded message -----
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by: To learn the basics of securing 
> your web site with SSL, click here to get a FREE TRIAL of a Thawte 
> Server Certificate: http://www.gothawte.com/rd524.html
> _______________________________________________
> agtk-devel mailing list
> agt...@li...
> https://lists.sourceforge.net/lists/listinfo/agtk-devel