See also: IRC log
We briefly verified which of the submissions of initial spec suggestions have materialised. Three items (Meta 1, Meta 2, Global 0) were re-assigned.
We started an in-depth discussion of the proposal by Felix and Ian for Links 1, Links 2, Core 8 and Links 3. Agreement was on an abstract level: We should be re-using tag and attribute names from existing specs where that makes sense, but without cluttering up our specification with unnecessary things -- we would include tags in our namespace that have names similar to existing specs where the meaning is similar, but avoid mixing namespaces where it is not needed. Regarding the specification of time, it was controversial whether we should allow for absolute time. As with everything, ease of use and interoperability must be weighed against expressive power of the language. Annotation of the semantic role should be separate from the link as such, as well as from the timing.
Felix and Ian will work out a new proposal on this basis, including concrete examples and an indication of mandatory and optional aspects, and send it to the public list.
All contributions as defined in the minutes of the previous meeting have materialised except for Meta 1, Meta 2 and Global 0
Marc: so I suggest to re-distribute these.
volunteers?
Marc: I can do meta 1
Enrico volunteers for meta 2
Felix volunteers for Global 0
scribe: by 15 August
<scribe> ACTION: Marc to specify Meta 1 - due 15 August [recorded in http://www.w3.org/2008/07/31-emotion-minutes.html#action01]
<trackbot> Created ACTION-28 - specify Meta 1 [on Marc Schröder - due 2008-08-15].
<scribe> ACTION: Enrico to specify Meta 2 - due 15 August [recorded in http://www.w3.org/2008/07/31-emotion-minutes.html#action02]
<trackbot> Created ACTION-29 - specify Meta 2 [on Enrico Zovato - due 2008-08-15].
<scribe> ACTION: Felix to specify Global 0 - due 15 August [recorded in http://www.w3.org/2008/07/31-emotion-minutes.html#action03]
<trackbot> Created ACTION-30 - specify Global 0 [on Felix Burkhardt - due 2008-08-15].
next item
Felix and Ian present the proposal they had circulated by email, for requirements Links 1, Links 2, Core 8 and Links 3.
<felix> General remarks:
<felix> The main question for things that are not directly related to emotion/affect is: shall we specify them ourself or advise the use of similar xml specifications (make or buy)?
<felix> arguments to use existing specifications:
<felix> It is not core to Emotion, so not worth us inventing something new
<felix> Many specifications exist
<felix> Using a format from one of our "neighbors" will help foster closer integration of our groups / standards
<Ian> http://lists.w3.org/Archives/Public/public-xg-emotion/2008Jul/att-0006/00-part
<felix> html is a ML to format docs in browsers [2].
<felix> SMIL (Synchronized Multimedia Integration Language) smil is a markup to format streaming media shows [3].
<felix> EMMA (Extensible MultiModal Annotation Markup Language) emma is a markup describe multimodal interaction [4].
Felix:Any other markup languages we should look at?
Marc:I suggest to look at InkML for the sampling mechanism
Ian: suggests not to re-invent the wheel, but
use other elements where they exist and are not specific to emotion
... asks how commonly used SMIL or EMMA are
Paolo: describes use of EMMA with speech server
for semantic recognition results for call centers etc.
... but original idea was to use in multimodal environments, and only few
projects currently seem to do that.
Ian: asks what people think of using html:a element
Felix: states that the element has too many
attributes that we don't need
... so suggestion by Christian Peter to reuse the names of
elements/attributes from existing languages but not the actual
elements/namespaces.
Ian: so that becomes a third option, between adopting existing solutions and re-inventing our own.
<felix> <emotion> <category name="pleasure"/>
<felix> <media src="file:///mySound.pcm"/ type="audio/basic"> </emotion>
<Bjoern> agrees with "target" over src
MArc: asks if we can use "target" rather than "src" because it seems to make more sense
Paolo: if we reuse the element, we should also
reuse the attribute names
... if we invent our own element, then we are free to choose names
<Ian> i like URI
Paolo: in SSML they had "audio", now switch to "media", and the attribute can be called either "src" or "uri"
Marc: likes "uri"
Ian: is "uri" maybe a protected keyword? it seems odd that no-one uses that.
<Bjoern> Bjoern does not consider URI intuitive...
Ian: it seems we like Christian's idea
Björn: it seems we want to re-use things but not completely stick to them.
So we could use this as a guideline for the detailed discussion: being inspired by existing specs but only use what we really need.
Marc:Why do we need the mime type?
Paolo: it brings us the type of data that the URI points to
Felix: it will be very important for automatic processing, and everybody now has it
Marc:maybe that's true
Marc:agrees it adds value, as long as it stays optional
Felix: agrees
Ian: that brings up an important point: which things are mandatory and which are optional.
Marc:last time we tried this (with the HUMAINE EARL) we ended up having everything optional.
Ian: so every tag should have at least one mandatory attribute.
<Bjoern> we certainly have to leave several items optional, non-the-less...
Felix and Ian will work out a concrete example on that basis.
Can we move on to Links 2?
Felix presents.
Felix: again, we should take the best of EMMA and SMIL that is useful for us, but not take everything.
<felix> <utterance id="utt1" emma:start="1231314123"/>
<felix> <nounPhrase id="np1" emma:start="1231316123" emma:duration="3s"/>
<felix> <emoml:emotion emma:start="2s" emma:duration="5s">
<felix> <emoml:category set="everyday" name="pleasure"/>
<felix> <emoml:start emma:time-ref-uri="#utt1"
<felix> emma:time-ref-anchor-point="start"
<felix> emma:offset-to-start="2s" />
<felix> <emoml:end emma:time-ref-uri="#np1"
<felix> emma:time-ref-anchor-point="end"
<felix> emma:offset-to-start="0" />
<felix> </emoml:emotion>
<Bjoern> also prefers intuitive dates.
Marc:I find absolute time in milliseconds since 1 Jan 1970 unintuitive
Marc: definitely I would like to avoid mixing these with human-readable times.
Enrico: agrees
Felix: couldn't we just leave that to the user
Ian: my feeling is we do not need to refer to
dates
... but it should be fine to tie our annotation to a zero-point
... so we need only relative times, not absolute times.
... also, I think we shouldn't use "30s" etc., because it is not
machine-readable.
<Bjoern> thinks that absolute times might make sense for us in a surveillance application?
Paolo: in EMMA, there are both kinds of timing,
absolute and relative
... but in the end, having absolute timing was not actually useful for
emma.
<Ian> http://www.w3.org/TR/SMIL3/smil-timing.html#Timing-ClockValueSyntax
<Ian> 50:00:10.25 = 50 hours, 10 seconds and 250 milliseconds
Felix: I don't see the need to force our users
into a specific way of describing time
... it is not relevant for emotion
Marc: this is an example of a very general point: interoperability.
Dylan: for what I can see, we are fine with relative time, e.g. since the start of a video
<Bjoern> Bjoern sees applications for absolute time: broad scale media retrieval, surveillance, ...
Felix: emotional diary would need absolute time
Dylan: disagrees
Marc:The question to me is, how complicated do we want to make this?
Marc:We see with EMMA how complicated it becomes when we want to make it powerful.
Felix: time is running up
Ian: let's continue discussing via the mailing list.
Ian presents.
Ian: felix made the point that this is similar
to our discussion on categories.
... we may have too many custom sets if we allow people to bring in their own
categories of semantics of links.
... I would suggest to distinguish clearly between the actual link and its
semantics.
... they are conceptually distinct.
Marc:We need to think about the relation between the three requirements (can we have a time wihtout a link, etc.)
Ian: in a concrete example, which attributes are optional, which are mandatory, what depends on what etc.
Felix: concrete examples will be more complex
Where should this be discussed?
Felix: let's have this in the public list, this is the heart of the matter.
<Bjoern> agrees
Björn: suggests the f2f in Cannes is the time to discuss details,
Björn: lets try to raise all issues by then.
Marc:Let's continue by email.
This is scribe.perl Revision: 1.133 of Date: 2008/01/18 18:48:51 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Found Scribe: marc Inferring ScribeNick: marc Found ScribeNick: marc WARNING: No "Topic:" lines found. Default Present: +049892892aaaa, marc, Bjoern, +035321425aabb, Paolo_Baggia, Enrico_Zovato, Dylan, Felix_Burkhardt, Ian_Wilson Present: Dylan Ian Bjoern Paolo Enrico Felix Marc Regrets: Andy Jianhua Agenda: http://lists.w3.org/Archives/Public/public-xg-emotion/2008Jul/0007.html Got date from IRC log name: 31 Jul 2008 Guessing minutes URL: http://www.w3.org/2008/07/31-emotion-minutes.html People with action items: enrico felix marc WARNING: No "Topic: ..." lines found! Resulting HTML may have an empty (invalid) <ol>...</ol>. Explanation: "Topic: ..." lines are used to indicate the start of new discussion topics or agenda items, such as: <dbooth> Topic: Review of Amy's report[End of scribe.perl diagnostic output]