- From: Felix Sasaki <fsasaki@w3.org>
- Date: Wed, 7 Nov 2012 18:00:53 +0100
- To: public-multilingualweb-lt@w3.org
- Message-ID: <CAL58czpWub7aACG6evxM0A8awA_+d8EKstXJZuSvNasgOB9CuQ@mail.gmail.com>
Hi all,
I made some AI and issue clean up, see below, based on the f2f outcome. The
minutes of day two are also below. I may have made mistakes, please check.
ACTIONS
action-164 Create an RDF ontology for ITS, and make sure that we update the
ITS XSD to accomodate the RDF output created via ACTION-147
closed since there is nobody moving this forward. Sebastian Hellmann asked
Tadej, Maxime and I in an offline mail (sorry for disclosing this here,
Sebastian) but I don't see time for this.
action-179 Work with Olaf-Michael to help draft blurb for international
organizations
Open for a long time, closing now.
action-257 Propose extensible of note to support ITS type to XLIFF TC
action-259 Review names of new attributes for XLIFF<->ITS mapping to ensure
understandability and consistency
Assume that these two can be closed, otherwise let me know, David.
action-262 write note about allowed characters issue with help from
Mauricio and karl
Done
action-264 Follow up on domain list topic with DCU closed
Done
action-269 David Lewis to check with Phil what his preference was between
quality issue locally with inline and script-based stand off
I would suggest that if we don't get feedback by next week Monday, we drop
the inline version
action-273 Discuss casing issue with Lucy Software
Done
action-276 Send XLIFF 2.0 spec to Richard
I assume done - David?
action-280 Edit the spec to unique standoff markup closed
Not needed anymore given
http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Nov/0059.html
action-283 Felix to integrate dave's email about standoff mechanism for
provenance into spec
Dave, can you send me a pointer about this?
ISSUES
I closed below issues, please protest if this is wrong.
issue-22 Provenance and agents . Covered by editing action items related to
standoff and name of the data category. A question here: I think we agreed
on the f2f to rename the data category to just provenance and to change the
attribute names accordingly, e.g.
"transPersonRef"
to
"personRef"
Is this correct?
issue-29 Please comment the possible solutions for HTML+ITS2.0 to HTML (and |
embedded by | embedding) RDF. Done, but without ontology.
issue-50 The dir attribute in HTML5 does not have the same value as in ITS
(will be done via Felix' action-250)
issue-54 ITS MIME Type (will be done via Felix' action-251)
issue-57 Allow for absolute location paths (will be done via Felix'
action-282)
=============
DAY 2 MINUTES
=============
[1]W3C
[1] http://www.w3.org/
- DRAFT -
MLW-LT f2f day 2
02 Nov 2012
[2]Agenda
[2] http://www.w3.org/International/multilingualweb/lt/wiki/LyonNov2012#Friday_2nd_Nov:_MLW-LT_WG_meeting_agenda
See also: [3]IRC log
[3] http://www.w3.org/2012/11/02-mlw-lt-irc
Attendees
Present
Ankit, Bert, Clemens, Dave, David, Dom,
Fantasai_(dir/ruby), Felix, Fredrik, Karl,
Kojii_(dir/ruby), Leroy, Matthias, Milan, Moritz,
Mārcis, Naoto(remote), Norbert_(dir/ruby), Pablo,
Richard_(dir/ruby), SebastianSk, Tadej(remote),
Yves(remote), jirka, matthiasK, mauricio, mhellwig,
pablo, pedro, renatb
Regrets
Chair
felix
Scribe
DomJones, fsasaki, mhellwig, Milan
Contents
* [4]Topics
1. [5]agenda review
2. [6]meeting with i18n wg
3. [7]domain lower casing
4. [8]Marcis' commments (part 1)
5. [9]reconvene with Marcis' comments
6. [10]Planning for 2013
7. [11]action item and issue review
8. [12]XLIFF Mapping Meeting [Issue-55]
* [13]Summary of Action Items
__________________________________________________________
agenda review
<fsasaki> waiting for people to come
<DomJones> Scribe: DomJones
Felix: Discussing presentation to HTML WG, Frederick showed
example use-case in OKAPI. Large group, many issues, individual
feedback is more likely from HTML WG.
<fsasaki>
[14]http://www.w3.org/International/multilingualweb/lt/wiki/Lyo
nNov2012#Friday_2nd_Nov:_MLW-LT_WG_meeting_agenda
[14] http://www.w3.org/International/multilingualweb/lt/wiki/LyonNov2012#Friday_2nd_Nov:_MLW-LT_WG_meeting_agenda
Felix: Dave, Dom, Leroy have to leave at 1 which topics do we
need them for. Morning agenda (until 13.00) is fine, XLIFF
discussion will be held at 3pm. Implementors slot is from
2-3pm.
… Starting goToMeeting
Naoto and Tadej join us on the goToMeeting.
<fsasaki>
[15]http://www.w3.org/International/multilingualweb/lt/wiki/Lyo
nNov2012#Friday_2nd_Nov:_MLW-LT_WG_meeting_agenda
[15] http://www.w3.org/International/multilingualweb/lt/wiki/LyonNov2012#Friday_2nd_Nov:_MLW-LT_WG_meeting_agenda
… We just had a meeting with HTML group, no specific outcome
but connection made between the WG. 1st topic is ITS tool
discussion.
Dave Lewis: Background to this - came from MT confidence score
generalised out to other data cats. Solution applies to any
data cat. Some cat's will contain confidence, quality, disambig
which may be different for every element / span. Most likely
that they use one value for tool information. Large overhead
for replication. Main example was proposal for MT confidence
score.
… You may have default done with one tool other certain
sections done with other tool. Global rules therefore cannot be
used. We need with a seperate data cat or a certain mechanism.
Suggestion to use trick used for standoff markup, not a data
category, contains tool information referencing part of that
tool element aligned to data category.
… Allows element referencing across a document.
… Could be over-written by element further on in document.
Dave Lewis: Yves had proposed some text but in order to take
furthur we needed to look at its application to other data
cats. I have looked at the ITS tool text and give examples on
current proposal for relevent data cats.
… have done this for MT confidence and Text A Annotation. If we
use this mechanism as a general purpose mechanism which seems
to work fairly well.
… you end up with data cats which only have a local attribute
(such as MT score) this combined with top level references for
tool informaiton.
… looking at definition of text A annotation you end up with
nearly the exact same pattern.
… have received comments back from Marcis on MT confidence
score.
David F: Tools should be made mandatory on more data
categories. Loc Qual Score (Precis) should be a candidate for
mandatory application of this. The same for Text A Annotation.
David Lewis: Have reduced it right down to local selectors, not
applicable to global.
Felix: In your presentation you state "not define external
format" this is not clear in the draft. You just have a URI.
Dave Lewis: We're probably a bit too generalised when we talk
of having a score for Text A Annotation.This could have been
used for Disambig, Terminology, Domain. The way we phrase that
allows several different data categories where the score is not
different from the process it relates to.
… Could have general purpose score attribute.
… MT confidence score, disambig, domain, Terminology. Would
this need to be more open ended?
… feedback from Tadej
Tadej: One thing which would be good to have is relation of
each instance to a score of the data category it relates to.
You can parse up the tree and see which data cats are produced
by which tool. Same text by terminology tools and Text A tools
at the same time. So could we direct tool-info at every node?
Dave Lewis: That is what we were trying to avoid with tool-info
with a mechanism for global declaration. Which ITS data
category annotations it applies to.
<fsasaki> its:toolsRef="MTConfidence|file:///tools.xml#T1
Disambiguation|file:///tools.xml#T2"
… For the element you are applying the declaration to you are
saying all of the data categories in that element were
generated by a specific tool. Different disambig / tools need
to be applied element by element. Worst case scenerio is every
element being done by a different tool but we dont think this
is a common situation.
Felix: The example pasted above, is this what you mean?
<fsasaki> also, tadej, is that the functionality you need?
Dave Lewis: Yes, gives flexibility for possible declaration of
every markup. I was interested to hear the feedback from others
as to whether we need different annotations for text
annotations, domain and terminology.
Marcis: If you dont look-up all instances in a term base, but
use extraction method for term-candidates you have the
confidence. Further you can fine-tune processes based on the
confidence.
… allows users to decide precision and recall which allows fine
tuning of systems.
<tadej> fsasaki: this is expressive enough, but may be verbose
for content which was annotated for multiple data categories -
it boils down how easy it is to relate every its-ta-confidence
instance to the tool it was produced by, where there are many
tools in the mix
Dave Lewis: Had been starting to think about this for demo
systems. Enricher run over text inserts alot of annotation
which may well result in false+
… How much do we know about the processes applied to
annotations?
… thresholds need to be added.
tadej: One solution to avoid verbosity is annotation of the
tool at the top-level of the document. Produce one annotation
on the root applied to all elements below.
Dave Lewis: ITS tool essential does that but data category is
bound to particular tool. Mark-up addresses that, we're taking
that a step forward to text analysis annotation. Is this "at
risk"?
Felix: No, "at risks" means that feature is clearly defined. To
do that we have only three weeks left.
Dave Lewis: If we happy with how we operate ITS-tools we need
to look at how we insert these data cats for Text A Confidence
score and for MT confidence. 1-2-1 matching to data categories.
For complicated like disambig are there more than one
confidence score depending on entity, lexical mapping, etc. Do
we need to be more fine grained in the confidence score there?
… There is the overview, questions on wording etc, my feelings
are that it seems to work on those data categories and the
knock on effect of combining confidence scores into one data
cat.
… Im looking for people interested to give us feedback now. Im
happy to continue to editing these but looking for feedback
from Marcis, Tadej, David F, Ankit.
Felix: 3 weeks is tight. Lots of test suite work needed in this
period. Suggest all those interested in this to look into this
today (2nd Nov). We will discuss again on Monday and try to fix
it completely so the other timeline is not effected. If
something comes up on Monday we have another week but we need
feedback by Monday on this.
<fsasaki> tadej, it seems we lost you on gotomeeting
Dave Lewis: Suggestion has a few typos etc, can people look at
that. Example annotation provided, some what editorial but we
have some examples as to how it works with different data
categories.
<tadej> fsasaki: reconnecting - the audio suddenly went silent.
felix: This has an impact on test-cases, needs to be in the
test suite.
<fsasaki> tadej, would that be a mandatory for text analytics?
asking also because of test suite etc.
David L: We have it as a general mechanism would not make sense
with a number of other categories.
David F: Unless I know the value / profile of the score it
provides nothing.
Felix: Are there tools which produce this score out of human
annotation.
… where scores are provided based on reviewing but a human.
<tadej> fsasaki: what exactly are you referring to as
mandatory? the confidence score mechanism, or the tool
reference mechanism?
David F: Score is an orthogonal feature.
felix: For MT we have MT-confidence, what other data categories
tool would produce that?
<fsasaki> tadej, I meant whether the tool mechanism should be
mandataory for implementors of text analysis annotation
Pedro: Any LSP would produce score for themselves. In scenerios
client request quality audit on content we produce or by 3rd
party. Important point - before quality audit you set the
methodology otherwise the audit is not valid at all.
<fsasaki> that is, for implementors of a score for text
analytics
Pedro: different LSPs have different metrics based on revision,
type of errors, severity and generates a score.
David F: Without a methodology you cannot produce score. May be
better to call is "quality calculation score" etc.
Felix: Precis is currently at risk without this methodology.
Tadej: Should this be mandatory? I think that without knowing
what produced the output it is hard to say anything about the
score. Which scores are comparable is hard to identify.
Dave L: We were talking about having a url that points to the
info, its a url of an element within that process info element.
The q: we refer to this process info element without stating
what the schema is but we state what the element is. Difference
is having a url that points to anything vs. not defining the
schema. In XML its fine can point to external or internal
element. But in HTML we need to specify how the url references
a url in that script.
Felix: You could have a seperate script element for each
standoff item.
… pitty Yves cannot be on the phone. He has raised concerns:
anything possible, tool with element it in, or define a schema.
There is a drawback that you restrict people to xml processing,
what about the case of RDF or audio. Does everyone who needs
the element need XML?
Pedro: This can be used by a client where a ref is used Score
is normally a relative value. You say if the threshold is X and
whether the content can be part of the profile ref.
Dave L: SMT gives the case where you are indexing the training
data to diff MT engines. No way to classify that we understand
at the moment. You may end up defining a MT by a description of
the MT egine.
Pedro: Not many impls as its hard to get that score
automatically.
Dave L: Self-generating score only used for comparison between
the same engines.
<fsasaki> "Disambiguation|file:///tools.xml#T2" >
"Disambiguation|[16]http://enrycher.com/v1.2/language-en"
[16] http://enrycher.com/v1.2/language-en
Felix: Paste proposal into chat, from Yves, this URI itself is
just a URI, no further information, self-contained in the URL.
This tool is X, in Lang Y, in the URL where each tool can
create the tool itself. But in a large document this is the
list of annotation with URI = tool1, URI = tool2. Dont restrict
the URI being retrieved and XML removes this restriction.
Dave Lewis: Naoto has interest in this.
… any other comments else I'll take this on board, update text,
get feedback from Tadej and Yves. Try to update and send off
today (2nd Nov).
Tadej: I like felix's suggestion on URI encoding. All people
will not be able to encode in a common format but good to
provide best-practices. I will send Dave L some comments.
… raise another point: In Dave's proposal mechanism for Text A
Annotation can only be applied to ITS data cats and not non-ITS
data cats. Is this something we would like to open up?
Dave L: Not sure on that wording, really a scoping thing. Could
be applied to meta-tags in HTML but this stretches scope of
Impl. Suggest we delete that. Unless others have a specific
use-case.
Tadej: When not used on ITS elements meaning is undefined.
Dave L: Can you email me that and I'll add it to the document.
David F: 2mins. I need to fix logistics for XLIFF meeting. Does
everyone want to use it or is it a breakout?
Felix. Timing it needs to be 3pm.
… 3-4pm xliff mapping meeting
… may make sense to have everyone here to review action items.
Move this to 4pm.
… updated agenda
[17]http://www.w3.org/International/multilingualweb/lt/wiki/Lyo
nNov2012
[17] http://www.w3.org/International/multilingualweb/lt/wiki/LyonNov2012
… Tadej / Noato will you join us this afternoon?
… Tadej no.
Felix: Propose we adjourn at 3pm and the XLIFF meeting can
follow.
meeting with i18n wg
<fsasaki> self-introduction of participants
<fsasaki>
[18]http://lists.w3.org/Archives/Public/public-i18n-core/2012Oc
tDec/0022.html
[18] http://lists.w3.org/Archives/Public/public-i18n-core/2012OctDec/0022.html
<fsasaki>
[19]http://lists.w3.org/Archives/Public/public-i18n-core/2012Oc
tDec/0022.html
[19] http://lists.w3.org/Archives/Public/public-i18n-core/2012OctDec/0022.html
Felix: HTML session introduced MLW-LT group. Would be good to
get feedback on a number of issues. Info share meeting with
L10n w3c group. 2 items are relevant for you. Directionality
and ruby information.
… values are given for directionality and ruby information.
… what is here is from the ITS 1.0 spec without changing
anything.
… Times have changed for directionality there are new
attributes, Ruby has a different ruby model than XHTML. So how
do we proceed?
… We are aiming to make people aware of what is possible for
directionality and ruby. Would be great to get your feedback.
We refer to what is being done for these 2 data cat in HTML5.
… For those using XML based examples the best thing would be
for them to use the HTML namespace. However if not possible
these elements could be defined in the ITS namespace.
… 1 other question: There is no rendering or processing here
involved which is hard for testing relating activities. Should
we just refer to these other places, would be good to get your
feedback.
r12a: Do we need to maintain backwards compat with ITS1.0.
Felix: Its not straightforward, a break may make sense. Not
sure it would break anything in content or applications. If we
need to break this backward compatibility we need to discuss
this in the group.
<fsasaki>
[20]http://www.w3.org/International/multilingualweb/lt/drafts/i
ts20/its20.html#directionality-definition (sec 6.5.1)
[20] http://www.w3.org/International/multilingualweb/lt/drafts/its20/its20.html#directionality-definition
r12a: ITS describes concepts that need to be supported for
internationalisation. Key thing: Express the concepts that need
to be supported in the markup. One thing you missed at the
HTML5 WG on bi-di, which you will not have heard.
… We started describing how to use HTML5 for bi-di. bdi element
and "auto" value ??
… they isolate certain text for dbi where you have text in HTML
and it interferes with stuff around it. Not only are problems
with dropping text into HTML but for bdi in general. Direction
can be assigned to text but can also isolate that text in
plaintext. People are encouraged to use those control codes as
opposed to existing methods.
… The CSS working group has retrofitted those ideas into the
CSS model. Looking for HTML WG to add two extra values to the
DIR attribute. Isolation is really important in bdi. Dir = LTR
/ RTL is to be avoided in replacement of new bdi attributes.
… proposed extension to HTML that would be retrofitted into
HTML5 during the CR phase (2014). Major shift, all fluid, many
questions remain.
Felix: Could we point to the HTML5 spec for directionality.
r12a: May not yet be in HTML5 by the time ITS2.0 is published.
Fantasai: Seems you have some values not already in HTML5.
Given this it makes sense to add values here, not worrying
about what HTML5 is doing. I dont think its a concern to sync
this feature with HTML5.
r12a: May be a problem as ITS 2.0 is looking to inform on how
bdi is used in HTML5.
Jirka: I think its no problem as we are providing mapping from
HTML model to our model. So its not too much of a problem to
add two new additional values to ITS.
… we can just extend our mapping from HTML5
Felix: People involved in XLIFF may have more information. At
LocWorld support of bi-direct support in XLIFF was discussed.
We are trying to copy the HTML5 model. That may be one area
where they may want more than guidance. They are near feature
freeze, David can you comment?
David F: Bi-direction support was added to draft. Feature
freeze informally before christmas / mid-january. Not trying to
mimic HTML. In XLIFF 1.2 unicode control chars were being used.
No Auto value in current draft, only LTR, RTL on structured or
inline elements.
… With have (in XLIFF) structural and in-line, not global and
local. They are not overlapping.
… current draft can be influenced. If this should be changed it
could. ITS to XLIFF mapping call today.
… important as its a major release, breaks backwards compat,
future releases (minor) will not change back-wards.
… No ness about attributes it would be about processing
requirements. Very little processing req. If you have input for
proc req then its the right time to influence the XLIFF group.
r12a: Likely to change. We have documentation on bdi. Inline
took line with minimum markup. New docs influence the way
people write bdi. Every word that changes is surrounded with
markup. Its a shift from previous approach.
David F: Even things which would be already considered very
local in XML are very structured in XLIFF.
Felix: Should we continue this now or table for 3pm? One aspect
to what Richard said from the beginning. ITS document provides
to people the right thing to do, therefore XLIFF people could
be directed to this.
r12a: isolate and automatically guess / assign directionality
are given by bdi. You can start a span of plaintext by LTR or
RTL.
David F: Is the auto approach a good idea to have in
localisation.
Norbert: There is an overlap between your work and my work.
There is aneed from ITS to work with Reg Exp with JavaScript.
Are there other req where you define things that would be
interpreted in Javascript.
<fsasaki>
[21]http://www.w3.org/International/multilingualweb/lt/drafts/i
ts20/its20.html#allowedchars-implementation
[21] http://www.w3.org/International/multilingualweb/lt/drafts/its20/its20.html#allowedchars-implementation
norbert: reg exp where ITS defines req exps interpreted by
Javascript. WE have to improve unicode support in reg exp so
your functionality would work. Are there other features of ITS
that need support in Javascript.
… you may be relying on other features which is not yet
supported.
… if nothing comes to mind right now we are also looking for
future input.
Felix: This is the main case where this may be used in
Javascript. I dont see it so much in other data cats where this
may be applied.
<fantasai> RRSAgent: pointer
Felix: ITS 2.0 moves to LC at end of Nov. I will send this to
you guys for review as to whether you think this is the right
way to be phrased. I need to talk to W3C about back-wards
compat with Directionality and Ruby. ITN group is busy but a
heads-up another call will be coming in Nov. We'll take it from
there.
r12a: What does MLW-LT think about bdi and ruby /
directionality.
Jirka: Im worried that the HTML spec was changed recently and
this has not been integrated into the spec yet. How to handle
more complex cases in Ruby etc. We should use same mark-up on
ruby as taken by the HTML but do we have time.
r12a: Ruby supoprt in 5.0 HTML, and isolation support. THe
problem is that 5.0 wont be finished before your spec if
finished.
Jirka: As long as ruby is stable in HTML 5.0 but I'm not sure
on this.
David F: Allowed to use normative references, are they in the
right state?
felix: We need to develop our testing and be part of our LC
draft. This doc provides guidance to do the right thing, rather
than having a normative definition.
David F: Data cats from 1.2 have moved from 1.0. If the
category is now in HTML would the right thing to say its no
longer in our scope as its in the HTML WG scope.
Felix: We still need to give guidance, albeit non normative
Jirka: Currently we try to copy what HTML is doing. What was in
ITS 1.0 we used XHTML base elements which were dropped. Would
be strange to add ruby in 2.0 to find it was later added to
HTML.
r12a: Brainstorming… In data cat world generic terms can be
described in prose. What currently being done in HTML5 in terms
of markup. Enables test in CR based on current HTML5 spec.
felix: Ruby tests are rendering based. We currently have no
browser / render based impls which means group cannot provide
the tests. People who provide normative usage are not in this
room. We need to agree upon this in the WG. We cannot get from
this group a normative and testable definition.
r12a: So this info should be in the spec but non normative.
felix: Yes, this also gives us more time. For example Nov 2013.
r12a: What would you say in this non-normative text.
felix: Currently state nothing but that this will be
back-filled in final draft. Provide placeholder for text and
move forward after LC. We could then work together to fill in
spec.
<fsasaki> scribe: fsasaki
dom: we have the opportunity to write what is happening next
year
... if we provide it non-normatively now
david: non-normative means that you don't use the words MUST,
SHOULD etc. and you don't need to do tests
richard: an application does not need to test things for
conformance
... you could not guarentee that XLIFF will have "placeholders"
for bidi stuff
<DomJones> scribe: DomJones
felix: Group based on EC funding and therefore time limited.
time extension are not a possibility which gives us a strict
timeline on this.
… any other thoughts from those here?
Fantasai: Asks for clarification, richard said your providing
recommendations on how mark-up should be applied to content.
felix: these recomendations are created based on inputs from
ITN working group. So others can look at how directionality /
ruby works.
… We're looking at how it works in HTML, guidance, not a
normative feature. Not replicating what is done normatively in
the HTML spec.
Fantasai: Looking at how to take HTML standards for
localisation and applies to other pieces of data. Would not
suggest using approach taken in HTML. 2 things: XHTML model and
current HTML model and not sure how it will look in future
models.
felix: Its a moving target
Fantasai: Should have one attribute for directionality. Not the
same as replacing with bits of HTML5.
felix: placeholder is a good agreement.
Jirka: Good to represent all values in directionality.
Felix: Who would test this normative features? Hoping we dont
define normatively as there are no test cases.
David F: Normative should be tabled for 2.5 or 2.1 ITS.
… they are unstable elsewhere so what can we actually do?
Fantasai: XML dir attribute with clear semantics. Have all RTL,
LTR, etc, all applied to one attribute. As opposed to multiple
attributes. Which maps to bdi algorithm using X and Y.
Felix: We can create such guidance. Is there someone from the
LTN group who would like to help us with this?
Fantasai: Aharon Lanin from google would be a good person for
this
Felix: And if he is not avliable?
r12a: Email us and we'll help you with this.
felix: Normative and non-normative (guidance) are our options.
r12a: From ITS conception we need to specify what information
is needed anywhere to support ruby and directionality.
Direction, isolate, RTL etc. This was applied in a number of
formats, DocBook etc. What I think Im hearing is we could do
this generic stuff but to get through CR phase you need to test
these things. If you cant map this, you can't test.
Felix: We have three weeks. Whether testable or not. Three
weeks to stable draft. As soon as its normative deadline is
three weeks away.
Jirka: Maybe go too deep into functionality.
Felix: If it is normative it is not done. You need to assure
rendering, impl.
Jirka: Displaying, rendering is a problem for styling.
Felix: Who here is implementing Directionality and Ruby?
Currently there are no testing provided for this. If its
normative you need an assertion that it is tested.
Jirka: Different case as it was in ITS 1.0, if you drop it you
miss backward compatability.
Fantasai: Are you defining technology or a spec for others or
guidance for others to define technology.
Felix: Except for Ruby and Directionality technology. Hence
proposing drop these.
... There are features we used to test Ruby and Directionality
in 1.0 which use XPATH not used in HTML.
Norbert: Why are we even talking about these if they are not
being used.
r12a: Important for spec but not being implemented.
... I would strongly support it being non-normative rather than
not having it there. Issue about stability as opposed to
whether it is need it or not.
Felix: Would it work if I re-draft current sections, send them
to you, with placeholders you can see and whether it makes
sense for LC draft? Would that be ok? At the actually LC we
have another opportunity to update.
<scribe> ACTION: on felix to draft the ruby and directionality
sections See
[22]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[23]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action01]
[22] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
<trackbot> Sorry, couldn't find on. You can review and register
nicknames at
<[24]http://www.w3.org/International/multilingualweb/lt/track/u
sers>.
[24] http://www.w3.org/International/multilingualweb/lt/track/users%3E.
<scribe> ACTION: on fsasaki to draft the ruby and
directionality sections See
[25]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[26]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action02]
[25] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
<trackbot> Sorry, couldn't find on. You can review and register
nicknames at
<[27]http://www.w3.org/International/multilingualweb/lt/track/u
sers>.
[27] http://www.w3.org/International/multilingualweb/lt/track/users%3E.
action on felix to draft the ruby and directionality sections
See [28]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
[28] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
<trackbot> Sorry, couldn't find on. You can review and register
nicknames at
<[29]http://www.w3.org/International/multilingualweb/lt/track/u
sers>.
[29] http://www.w3.org/International/multilingualweb/lt/track/users%3E.
<scribe> ACTION: on felix2 to draft the ruby and directionality
sections See
[30]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[31]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action03]
[30] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
<trackbot> Sorry, couldn't find on. You can review and register
nicknames at
<[32]http://www.w3.org/International/multilingualweb/lt/track/u
sers>.
[32] http://www.w3.org/International/multilingualweb/lt/track/users%3E.
<mhellwig> scribe: mhellwig
fsasaki reviewing agenda
domain lower casing
yves: what do we return the lowercase value or the original
value?
<scribe> ACTION: pablo to talk to Lucy about casing issue
[recorded in
[33]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action04]
<trackbot> Sorry, ambiguous username (more than one match) -
pablo
<trackbot> Try using a different identifier, such as family
name or username (eg. pnietoca, pbada)
<scribe> ACTION: paolo to discuss casing issue with Lucy
Software [recorded in
[34]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action05]
<trackbot> Sorry, couldn't find paolo. You can review and
register nicknames at
<[35]http://www.w3.org/International/multilingualweb/lt/track/u
sers>.
[35] http://www.w3.org/International/multilingualweb/lt/track/users%3E.
<pnietoca> ACTION: pnietoca to discuss casing issue with Lucy
Software [recorded in
[36]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action06]
<trackbot> Created ACTION-273 - Discuss casing issue with Lucy
Software [on Pablo Nieto Caride - due 2012-11-09].
fsasaki: domain pointers can have v. long XPATH expressions.
Absolute location paths would make it shorter.
jirka: write location path for XPATH. Allow absolute and relative locati
on paths.
fsasaki agrees
jirka: we should rewrite the specification completely.
action: fsasaki to edit specification to resolve location path issue
Marcis' comments (part 1)
Marcis: to analyse for terms, you have to break down the document
... you need to do the analysis several times, for different domains
... unguided term annotation: annotations are made with confidence score
s
... a second way is to recognise terms in term base and annotate. there
you don't have confidence. either you have the term in your term or you
don't
... question arises: how do we add tool info?
... and another question for group: what happens when external rules are
not available?
fsasaki: we have linked global rules. the conformance section we say tha
t systems must process these rules
dave: we haven't defined what happens if it breaks down. we suppose a 'b
est effort basis'
Marcis: translate data category defines what should be analysed. This do
es not exist for other data categories.
... do we need a definition?
dave: there isn't a definition. it hasn't come up. [in case of annotatio
ns] you just do it for the whole document.
... no new mechanism is needed. maybe we need a discussion about this
Marcis: for translation it's more critical. annotation you add, you don'
t replace anything
dave: you could have false positives and then there's cost for going thr
ough and cross these false positives out. it will just raise cost
Marcis: does global override local?
dave: yes
Marcis: terminology is not to be inheritable, but what about this case [
discusses example in his notes]
fsasaki: we are looking at an example with nested elements around which
there is a term annotation around it
... need additional item information for each element.
Marcis: would it be the same in disambig?
fsasaki: yes, disambig is not inherited
fsasaki: [to tadej] would enrycher support nested elements
tadej: it's possible, i don't see why it wouldn't be. we're safe here.
Marcis: but you cannot do it locally?
tadej, fsasaki: yes you can
Marcis: what about overlapping annotations
fsasaki: won't solve
Marcis: also agent and tool information. will not go into detail at this
point
... there are a lot of very finegrained usages in agent provenance
dave: we now have a standoff mechanism, so does it makes sense to have a
tool which says provenance type = ???
... which would make the metadata definition easier
... so i tihnk it's a good idea to implement a standoff mechanism
action: dave to write an email to fsasaki who will integrate this into t
he spec
action: fsasaki to integrate dave's email about standoff mechanism for p
rovenance into spec
reconvene with Marcis' comments
<fsasaki> Marcis: language will fall back to language "english"
as a fallback
<fsasaki> .. in MT it is important that you know to which
language you are translating
<fsasaki> Ankit: difference between language is not ideal
<fsasaki> David: not an issue of ITS
<fsasaki> Marcis: sure, just a comment
<fsasaki> David: any industry implementation does mapping
anyway
<fsasaki> .. mappings are possible, e.g. to map any English
into your English
<fsasaki> Marcis: yes, like reading the 1st two characters
<fsasaki> David: yes
Planning for 2013
fsasaki: we need to do some event planning
fsasaki shows dates and events listed in excel document
fsasaki: F2F in January, workshop in Rome [March]
pedro: going to Gala.
fsasaki: next up F2F in april. tadej, is that good for you? everybody, p
leaes check your calender
tadej: I checked with hotel, availability end of april and a few times b
eginning of May
... May better to climb ...??
fsasaki: would 7th and 8th of May [$1\47] work?
agreement from group
... great. let's have the F2F meeting then.
... what about locworld. anybody going?
dfilip: we are thinking about FEISGILTT in London as we had good attract
ion and follow-up in Seattle so London should work well.
pedro: will probably go and Lucy Software will also be there (at locworl
d)
... I think we should submit as much as we can. not just be there
dfilip: we may have entry to the main programme through feisgiltt.
pedro: most important that we showcases are already running, we have to
show them
fsasaki: any other events to showcase
pedro: I propose an F2F in Madrid in July
... I'll check if the university is available
fsasaki: any other events you may present
mhellwig: DrupalCon at the end of March
fsasaki: [to dave] can you present at XML prague?
dave: ???
dave: also, world wide web conference. submission deadline soon. 13th-17
th May 2013
ankit: September 2013 I'll go to MT summit user track
dfilip: LRC conference. Around 20th of september 2013
fsasaki: time to close. anything else? when is the drupalcon
mhellwig: there's two. we'll go to one at least
fsasaki: where should the final event be?
jirka: when is it supposed to be. Oct, Nov, Dec 2013?
fsasaki: yes around there, depending of location availability etc. it's
supposed to be our largest workshow. We have a lot of implementations al
ready, so the critical one will be Rome.
action item and issue review
<fsasaki>
[37]http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action03
[37] http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action03
<fsasaki>
[38]https://www.w3.org/International/multilingualweb/lt/track/a
ctions/open
[38] https://www.w3.org/International/multilingualweb/lt/track/actions/open
<fsasaki> close action-231
<trackbot> ACTION-231 Create tests for its:param closed
<fsasaki> close action-255
<trackbot> ACTION-255 Determine and correct wording for
ISSUE-34 closed
<fsasaki> close action-258
<trackbot> ACTION-258 Ask XLIFF TC what best practice of
mapping ITS into a namespace in XLIFF closed
<fsasaki> action-268: see
[39]http://lists.w3.org/Archives/Public/public-multilingualweb-
lt-tests/2012Nov/0000.html
[39] http://lists.w3.org/Archives/Public/public-multilingualweb-lt-tests/2012Nov/0000.html
<trackbot> ACTION-268 Make sure that schedule for test suite
and schema update discussed at
[40]http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken
into account notes added
[40] http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30
<fsasaki> close action-268
<trackbot> ACTION-268 Make sure that schedule for test suite
and schema update discussed at
[41]http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken
into account closed
[41] http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30
<fsasaki>
[42]http://www.timeanddate.com/worldclock/fixedtime.html?iso=20
121105T15
[42] http://www.timeanddate.com/worldclock/fixedtime.html?iso=20121105T15
<fsasaki> ACTION: felix to send info about call time [recorded
in
[43]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action07]
<trackbot> Created ACTION-274 - Send info about call time [on
Felix Sasaki - due 2012-11-09].
<fsasaki> action-270: done via
[44]http://lists.w3.org/Archives/Public/public-multilingualweb-
lt/2012Nov/0009.html
[44] http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Nov/0009.html
<trackbot> ACTION-270 Ask phil and des and arle about need and
implementation committment for localization precis during next
call notes added
<fsasaki> close action-270
<trackbot> ACTION-270 Ask phil and des and arle about need and
implementation committment for localization precis during next
call closed
<fsasaki> action-271: dublicate of action-273
<trackbot> ACTION-271 Add a step regarding the lowercasing of
the domain data category notes added
<fsasaki> close action-271
<trackbot> ACTION-271 Add a step regarding the lowercasing of
the domain data category closed
<fsasaki> close issue-52
<trackbot> ISSUE-52 Domain in HTML5 closed
<fsasaki> "[Ed. note: Following schema example has to updated
once we have final XSD schema for ITS 2.0]" - drop example and
note
<fsasaki> "[Ed. note: All selector related definitions has to
be update to reflect queryLanguage]" - some data category
definitions refer to XPath expressions; need to generalize that
to refer to "relative or absolute selector"
<fsasaki> "[Ed. note: Need to reevaluate above statement
related to ODF.]" - remove paragraph above the note, that's it
<fsasaki> "The entity type follows inheritance rules." - delete
the sentence? came back to Tadej
<fsasaki> "[Ed. note: Below note is taken from the quality
issue data category. ..." - can be deleted
<fsasaki> "[Ed. note: Should locQualityIssues also be defined
for global rules? It seems not to be specific to local.]" - not
decided yet
<fsasaki> yves: having a generic container that is nice
<fsasaki> ACTION: yves to summarized "one container name"
proposal again [recorded in
[45]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action08]
<trackbot> Created ACTION-275 - Summarized "one container name"
proposal again [on Yves Savourel - due 2012-11-09].
<fsasaki> "[Ed. note: Missing the local mtconfidencescore
attribute.]" - to be done after or during tool definition
update
XLIFF Mapping Meeting [Issue-55]
<dF> Scribe: Milan
<dF> Chair: dF
Richard and Koji are with us, for bidi and Ruby to discuss
<Yves_> see also section on bidid in draft of XLIFF 2.0
[46]https://tools.oasis-open.org/version-control/browse/wsvn/xl
iff/trunk/xliff-20/xliff-core.pdf
[46] https://tools.oasis-open.org/version-control/browse/wsvn/xliff/trunk/xliff-20/xliff-core.pdf
Most of implementations are in XLIFF 1.2, version 2.0 is
currently under construction
Mappings are similar (structurally)
Let's start with Directionality (then Ruby)
dF: Inline doesn't feature to cover those
... XLIFF proposal for directionality in 2.0
Yves_: Any inline element (including <mrk>) has attribute for
directionality
<Yves_> See Bidi section here:
[47]https://tools.oasis-open.org/version-control/browse/wsvn/xl
iff/trunk/xliff-20/xliff-core.pdf
[47] https://tools.oasis-open.org/version-control/browse/wsvn/xliff/trunk/xliff-20/xliff-core.pdf
dF: Masking vs. <mrk> - explaining difference
r12a: HTML5 includes bdi attribute provides isolation mechanism
... HTML WG to provide a new value (Auto), decided
directionality based on first strong character
<dF>
[48]https://tools.oasis-open.org/version-control/browse/wsvn/xl
iff/trunk/xliff-20/xliff-core.pdf
[48] https://tools.oasis-open.org/version-control/browse/wsvn/xliff/trunk/xliff-20/xliff-core.pdf
<scribe> ACTION: dF to send XLIFF 2.0 spec to Richard [recorded
in
[49]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action09]
<trackbot> Created ACTION-276 - Send XLIFF 2.0 spec to Richard
[on David Filip - due 2012-11-09].
dF: There was never mechanicsm like Ruby in XLIFF
... can be provided as a context
... fs can help(?)
... XLIFF is a transport format, not resolved displaying
issues. Depends on tools how the content is displayed
Continuing the XLIFF Maping Table (r12a and Koji left)
Translation Agent Provenance skipped, not Dave
Text Analysis Annotation skipped
Target Pointer drives an extraction, there is nothing to
represent
Id Value as a resname in 1.2, no equivalent in 2.0
dF: Yves to propose rename on unit in XLIFF 2.0
... it doesn't have any sense to have ID value for inlines
(remove questionmarks)
Preserve Space solved at segment level (xml:space) but not for
inline
dF: could be used in sub-flow
Localization Quality Issue, hold till call with XLIFF committee
at Nov 6th
Localization Quality Précis
dF: We need a mechanism to reference an Agent
... who provided quality check
MT Confidence
Allowed Characters
dF: Do we need it for inline?
Yves_: Yes, example might be Login name restriction
Storage Size, issue only in 2.0
dF: push harder to have <mrk> extensible
... We stabilized what was possible
<scribe> ACTION: dF To color-code cells in Mappings table
dependent on unstable ITS categories or in XLIFF [recorded in
[50]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action10]
<trackbot> Created ACTION-277 - Color-code cells in Mappings
table dependent on unstable ITS categories or in XLIFF [on
David Filip - due 2012-11-09].
Summary of Action Items
[NEW] ACTION: dF To color-code cells in Mappings table
dependent on unstable ITS categories or in XLIFF [recorded in
[51]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action10]
[NEW] ACTION: dF to send XLIFF 2.0 spec to Richard [recorded in
[52]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action09]
[NEW] ACTION: felix to send info about call time [recorded in
[53]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action07]
[NEW] ACTION: on felix to draft the ruby and directionality
sections See
[54]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[55]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action01]
[NEW] ACTION: on felix2 to draft the ruby and directionality
sections See
[56]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[57]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action03]
[NEW] ACTION: on fsasaki to draft the ruby and directionality
sections See
[58]http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29 recorded
in
[59]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action02]
[NEW] ACTION: pablo to talk to Lucy about casing issue
[recorded in
[60]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action04]
[NEW] ACTION: paolo to discuss casing issue with Lucy Software
[recorded in
[61]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action05]
[NEW] ACTION: pnietoca to discuss casing issue with Lucy
Software [recorded in
[62]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action06]
[NEW] ACTION: yves to summarized "one container name" proposal
again [recorded in
[63]http://www.w3.org/2012/11/02-mlw-lt-minutes.html#action08]
[54] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
[56] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
[58] http://www.w3.org/2012/11/02-mlw-lt-irc#T11-14-29
[End of minutes]
__________________________________________________________
Minutes formatted by David Booth's [64]scribe.perl version
1.137 ([65]CVS log)
$Date: 2012/11/07 16:15:17 $
[64] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm
[65] http://dev.w3.org/cvsweb/2002/scribe/
Received on Wednesday, 7 November 2012 17:01:47 UTC