W3C home > Mailing lists > Public > public-swbp-wg@w3.org > September 2005

Re: [ALL/MM] EMMA Last call

From: Max Froumentin <mf@w3.org>
Date: Fri, 30 Sep 2005 16:32:54 +0100
To: Dan Brickley <danbri@w3.org>
Cc: Jacco van Ossenbruggen <Jacco.van.Ossenbruggen@cwi.nl>, swbp <public-swbp-wg@w3.org>
Message-ID: <874q82bmft.fsf@w3.org>

Dan Brickley <danbri@w3.org> writes:

>> I just bumbed into this and was wondering if this working group has,
>> or needs to have, an opinion about the Last Call EMMA draft at
>>
>>    http://www.w3.org/TR/emma/
>>
>> The language is described as a language for annotation and metadata.
>> Personally, I would have expected it to build on RDF, but it doesn't.
>> In fact, the word RDF isn't mentioned at all, while it spends quite
>> some sentences on its (rather complicated) relationship with XML. Is
>> it reasonable to ask them to clarify the relationship with RDF/OWL
>> too?
>
>
> I spent some time talking with Max (cc:'d) earlier in the design. RDF 
> was certainly
> considered. My understanding was that the need to attach probabilistic 
> and other
> metadata at a per-property level, made usage of RDF unwieldy (rdf's current
> reification vocab is just no practical there). Max, perhaps you could 
> expand? Is there
> a pointer to earlier RDF-ish designs somewhere?

Oh yes, such fun we had that summer, talking so much about EMMA with
danbri and dajobe...

I can't find anything easily, apart from the previous drafts with RDF
syntax until it was dropped
(e.g. <http://www.w3.org/TR/2003/WD-emma-20031218/>). The problems
which lead to a non-RDF syntax were of a practical nature and IIRC
were along the following lines:

- Refification, obviously. EMMA expresses things like "the probability
  that this utterance means "Bristol" is 75%.

- because EMMA documents are transient messages (between a recogniser
  and an interpreter, they don't necessarily have a URI and wouldn't
  typically be registered in a triple store, or mixed with other namespaces.
  So that advantage of RDF wasn't very valuable to us.

- Labelling recognition markup in XML with emma properties (like
  emma:timestamp) in RDF forced us to split an EMMA instance in two.
  See example in
  <http://www.w3.org/TR/2003/WD-emma-20031218/#s2.1.3.2>. This made
  parsing more complex, especially if the RDF/XML wasn't constrained.
  It was also necessary to introduce an XPointer scheme (see
  <http://www.w3.org/TR/2003/WD-emma-20031218/#s2.1.2>) which was
  deemed too complex for processing on embedded systems, where
  EMMA typically resides. 

Hope this helps.

Max.
Received on Friday, 30 September 2005 15:32:04 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:31:12 UTC