MLW-LT f2f -- 23 Jan 2013

roll call

<fsasaki> checking attendance ...

http://tinyurl.com/its20-comments-handling

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/PragueJan2013f2f#Agenda

<daveL> scribe daveL

issue-67

<daveL> yves: had no feedback from shaun to date so we probably can't advance here

<daveL> felix: comment could be addressed by dropping the ref to XML schema

<daveL> yves: will respond on issue 105

issue-69

<pnietoca> External rules may also have links to other external rules (see example 20). The linking mechanism is recursive, and subsequently after the processing the rules MUST be read top-down (see example 21).

<fsasaki> http://www.w3.org/TR/2012/WD-its20-20121206/#link-external-rules

<daveL> pablo: had responded that this was clear in the specification, but suggest a clarification

<pnietoca> the section is 5.4. (last paragraph)

<daveL> felix: confirms this is just a clarification

<pnietoca> change it

<fsasaki> "The linking mechanism is recursive" > "The linking mechanism is recursive in a depth-first approach"

<daveL> tadej: perhaps explain this recursion as being 'depth first' to be understandable more by computer scientists

issue-70

<daveL> felix: ref to section 5.5

<fsasaki> http://www.w3.org/TR/2012/WD-its20-20121206/#selection-precedence

<fsasaki> will add one entry between "global selections" and "data category defaults" for inherited information, but not specific to local markup

issue-71

<fsasaki> annotatorsRef

<fsasaki> scribe: fsasaki

daveL: Yves said the problem is: you can have a lot of annotatorRefs
... issue is: how to deal with annotatorRefs with two instances of local standoff markup
... e.g. lq localization issues and provenacne records
... so you can have multiple records of the same data category applying to the same selection
... you don't get the information whether the information comes from different processes
... Yves suggested whether we can put the information into the same ...
... my view was: for provenacne annotator ref is not that important
... so in the mail last night: could we exlude the lqi and provenance from annotatorsRef
... annotatorRefs is telling you what provided the provenacne annotation

tadej: from provenance it is not needed, but for lqi?

dave: don't think so for lqissue.

yves: sounds weird: have annotatorsRef mandatory for some data cats, possible for others, forbidden for two ...
... currently it is required for mt-confidence and disambiguation

<Marcis> ... and Terminology

yves: otehr solution: you could have it mandatory for these two data categories, and don't have it for others
... that would make things a lot simpler

dave: agree - not having two features interacting (standoff and annotatorsRef) would be good

felix potential resolution - so keep it mandatory for mt-confidence, disambiguation and term, and edit the list of data category items in the spec

<scribe> scribe: daveL

<fsasaki> ACTION: dLewis6 to come back to chase and kevin about discussion of issue-71 http://www.w3.org/2013/01/23-mlw-lt-irc#T08-34-49 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action01]

<trackbot> Created ACTION-388 - Come back to chase and kevin about discussion of issue-71 http://www.w3.org/2013/01/23-mlw-lt-irc#T08-34-49 [on David Lewis - due 2013-01-30].

<fsasaki> ACTION: felix to change example http://www.w3.org/TR/2012/WD-its20-20121206/#EX-its-tool-annotation-1 if the agree on issue-71 , see discussion at http://www.w3.org/2013/01/23-mlw-lt-irc#T08-34-49 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action02]

<trackbot> Created ACTION-389 - Change example http://www.w3.org/TR/2012/WD-its20-20121206/#EX-its-tool-annotation-1 if the agree on issue-71 , see discussion at http://www.w3.org/2013/01/23-mlw-lt-irc#T08-34-49 [on Felix Sasaki - due 2013-01-30].

felix: example 28 needs to be revised also, will do this now

<fsasaki> scribe: fsasaki

daveL: using the example in the test file - should we have usage of the data categories in the elements?

yves: yes

<daveL> dave: this example doesn't actually include the data category attributes to which the annotatorRef refers

<daveL> felix: makes note that the test file and the example should be revised to include this

yves: we don't have annotatorsRef for all disambiguation examples

<daveL> yves: we don't have annotatorRef in all examples of disambiguation

<scribe> ACTION: tadej to check disambiguation examples with regards to presence of annotatorsRef [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action03]

<trackbot> Created ACTION-390 - Check disambiguation examples with regards to presence of annotatorsRef [on Tadej Štajner - due 2013-01-30].

ISSUE-72 NIF comment

<daveL> felix: comment was which version of NIF do we refer to

http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0015.html

<daveL> .. there are 1.0 and 2.0

<daveL> .. also there stabilit was raises

<daveL> ... and Christian also raised whether the mapping was canonical

<daveL> dF: it may be a useful clarification for implementators

<daveL> felix: but its not clear what is meant by 'canonical XML' in this case

<daveL> tadej: it implied there should be a canonical XML serialisation

<daveL> felix: would such a requirement raise a bar for implementors, this need to be dicussed further on the lists

<daveL> felix: now will attempt to dial in Christian

issue-68

<scribe> scribe: fsasaki

marcis: there was a discussion on ITS term and disambiguation
... christian brought it up, various comments from the WG
... david suggested that we should not break ITS1.0, but felix said it is not necessary to have it

<daveL> marcis: summarises discussion

<daveL> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0152.html

daveF: don't break it if it works
... that's the bottom line
... we want to keep also independence of features

marcis: I could implement terminology independent of the rest of disambiugation
... the question is: if we agree to change something, it is independent, so different question
... david suggested to have a bp document that specifies how things relate

daveF: there are seperate use cases for disambiguation and terminology
... things are backed by different use cases, also from the implementers point of view

felix: we can also depcreate one of these

tadej: if we want to annotate the same fragment - which one to choose?

marcis: that is the biggest problem
... we cannot do both
... there was a comment from yves, we should break larger problems into smaller ones
... so even if we have an "upper level" data category which we could then use for both scenarios

tadej: we could use the same trip we did with annotators ref, e.g. using multiple values in the same attribute
... not sure if we would encourage people to do this
... complex, but same level of complexity as ...
... another solution tadej suggested was to have many attributes , but that's the same as having everything in one attribute
... if we can come up with a closed set of types of annotation, that's a solution
... but that needs to be a closed set, since we are specifying attributes
... right now for disambiguation we agreed for three levels: concept, entity, lexicon

marcis: there is no definition for each of these levels, e.g. what is a lexical concept?
... I saw that there is a terminology inconcistency
... terminology is not used always in the same way in the disambiguation description

daveL: the issue in using both of them for the same term - we are not clear how to combine them?

tadej: it is not an issue at the moment
... if you fold it in one data category, it becomes a problem

<chriLi> queue

daveF: a big system will have a terminology life cycle with many manual people, but it is an automatic workflow

daveL: aim of disambiguation is that it would make the output of automatic annotation available

christian: thanks to marcis for putting everything into a condensed form
... there are we with the discussion today: my understanding is the following:
... people think it is not a bad idea to try to come up with a data category that can subsume what ITS2 terminology and ITS2 disambiguation try to cover
... with respect to paying attention to ITS1: situation is that there is no need to go for backwards compatibility
... one way to achieve soft transition would be to deprecate existing ITS term
... one way to come up with the upper level data category: two implementation suggestions were made: based on attrbiute values and distinct values for annotation types
... this is how I understand the current state of the discussion
... I'm wondering what the next step would be
... to say: we realize that we want to really look into this change
... and want to do something to the current draft
... if this wants to be driven it could be done via mail or a seperate call
... need to agree on the approach

<daveL> scrie: daveL

<daveL> scribe: daveL

felix: we have agreement that backward compatability isn't an absolute barrier
... but it is in my view desirable

Christian: fully agree

felix: another point is trying in general to reduce level of substantive change
... another point is experience of people who implement and knwo users of its1.0 terminology
... such as yves and OKAPI community

yves: not necessarily a big problem to change but would like to keep backward compatibility in general

tadej: suggested changes would break backward compatibility

macis: potetnially we add complexity to terminology by including link to external ontology or other lexical resource

df: agrees

felix: compromise is having an umbrella data category, and allow term to stay the same

<fsasaki> arle: agree with marcis

marcis: have some questionns about the definition of disambiguation, e.g. the meaning of what is a lexical concept

christian: support having an umbrella data category that would not increase complexity of seaprate term and disambiguation use case
... also we will get better uptake if we can offer an easier route to marking up the output of text analysis
... rather than having to support the more complex issues in disambiguation

tadej: the reason for defining granularities was the major requirements of linguists, it was not sufficient to have this all in the target external data structure
... so even granularity definition was a compromise

arle: the term 'granularity' may also be an issue

tadej: was previously 'disambiguation type', but it was difficult to find the right term

felxi: asks tadej, marcis, christan to come up with a proposal that allows for both use cases and consider backward comatibility for term?
... but this would need to be done by the end of next week?

<Arle> Without putting too much thought into it, would disambiguationClassType work? Would this always correspond to a description of the kind of disambiguationClass intended?

christian: happy to let marcis and tadej to try and draft something over these two days and then I can dial in again to discuss it further

marcis: asks who was originator of disambig

tadej: originally it was a named entity recoginiser category, but after discussion also became merged with diasambiguation afteter discussion with linguasev and others

marcis: could we have a cascading model, since named entity can be composite

<chriLi> Don't forget to bring the beer bottles to the room as well :-)

daveL: note this overlaps with issue-109 on disambiguation in indic languages

issue-75

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0143.html

felix: jorge as shepard has produced a summary of this topic

christian: my domain comment had three parts
... one main point - was looking for a way for providing to meta-data on a domain without pointing to resource, this has no eyyt been resolved
... another point was that domain meta-data is processor specific
... so in one world it is called x then the context in which x is meaningful needs to be provided

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0136.html

christian: now jorge has resolved point 2b, but the baove has still also to be resolved

felix: felt adding this context meta was a new feature but could be reolved with a note that this relates to a single engine use case

<fsasaki> https://www.w3.org/International/multilingualweb/lt/track/products/9

christian; broadly agrees such a note would satisfy him, since in ITS the focus was on scenarios with a single engine scenario. But this need to be made clear as an assumption in ITS2.0

felix: have now started collacting items on tracker categories as 'not addressed in ITS2.0'

issue-73

felix: so if larger implementors, e.g. sap, adobe, ms, will but resoruces into the multiengine scenario we could consider it, other we should stick with making explicit the single engine context

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0015.html

felix: with NIF the stability is an issue and will refer back to sebastian Helleman about the plan for this
... need this information to react fully to this comment
... other comment was how the mapping could benefit from canonical definiition of mapping
... so my comment is whether this would be of use to implementors, since in the room there was a lot of familiarisation with the use and benefits of canonicalisation

christian: asks do we have more than one implementation

felix: confirms we have one from sebastian and one from felix

christian: I brought this up to ensure that whenever NIF processing is ensured, we end up with the same representation, and this needs normalisation and canonicalisation
... if not, then we may end up with versions that are incompatible

felix: asks whether some comparison between document in NIF is an likely use case. would the comparison not takeplace back in the document itself

christian: I think you would need a unicode normalisation

felix: but this was related to regex in another data category

christian: if we are reocmmending normalisation anyway in this other data category, could we not use this to solve the problem here

issue-74

<fsasaki> scribe: fsasaki

daveL: christian provided some bullet point comments
... are you planning more re-writing
... or should david and I take your comments in?

christian: if it would be ok with you
... I could turn the bullet points that people could read
... with respect with the general approach
... I could do editing of the doc
... by mid next week

https://www.w3.org/International/multilingualweb/lt/track/actions/377

that would be action-377

davidF: that's clarificatory stutt, not very urgent
... will wait for christian for a more readable version

felix: so we have discusesed all comments from christian

felix wil put thoughts on NIF in a mail

<scribe> scribe: Yves_

<scribe> Scribe: Yves_

issue-72

<fsasaki> original comment here http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0013.html

<fsasaki> .. see "Section 8.12 (Provenance Data Category)"

daveL: Provenance issue is about timestamp
... quite complex to implement
... e.g when the information is capture, etc.
... This is covered by the PROV standard
... and we have a mechanism to point to that
... so no need in ITS

<fsasaki> yves: so has the order of provenance a meaning?

daveL: so order SHOULD reflect the order things were added in the document

original commentor got a reply and we are waiting for a response. comment was rejected.

issue-76

Arle: need to re-look at it

issue-77

Jirka: proposal for a solution is in the issue's note.
... question was about HTML and rules precedence

Jirka: no need to change anything
... link is the same as link in global rules

<fsasaki> resolution proposal - see note from jirka Kosek, 22 Jan 2013, 22:58:35 at http://www.w3.org/International/multilingualweb/lt/track/issues/77

Marcis: my comment was that it was difficult to understand how things work
... because it's defined in multiple places

felix: in section 6.4 there are some explanation
... we would add Jirka's clarification there
... this would define the inheritance behavior

jirka: maybe issue is that global rules need to be read in document order

<fsasaki> "Global selections in documents (using mechanism of external global rules or inline global rules)" > "Global selections in documents (using mechanism of external global rules or inline global rules), to be processed in document order"

Felix: could point to 5.2.1 in the HTML section
... let's close this issue. See the note in the issue page.

<fsasaki> ACTION: jirka to make edit for issue-77 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action04]

<trackbot> Created ACTION-391 - Make edit for issue-77 [on Jirka Kosek - due 2013-01-30].

issue-76 again

Arle: an implementer was looking at issue's type
... and saw inconsistency

<fsasaki> original comment at http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0026.html

Arle: solution would be to change the definition
... add "or text is translated inconsistently"
... and a second example.

<Arle> Proposed change: The text is inconsistent within itself or text is translated inconsistently (NB: not for use with terminology inconsistency).

<Arle> Add second example: The translated text uses different wording for a single regulatory notice in the source that occurs multiple times in a series of manuals.

<fsasaki> change in this sec http://www.w3.org/TR/2012/WD-its20-20121206/#lqissue-typevalues

<scribe> ACTION: arle to make the edit for issue 76 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action05]

<trackbot> Created ACTION-392 - Make the edit for issue 76 [on Arle Lommel - due 2013-01-30].

issue-78

Felix: rel-type was registered, no more action is needed.

Felix: wrote a reply to that comment

issue-79

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0145.html

Felix: added text indicating namespace prefix can be difference than its if it exists already

Jirka: this just duplicate information. not good
... the initial text should already address the comment

<fsasaki> "The namespace URI that MUST be used by implementations of this specification is:" > "The namespace URI that MUST be used by XML-based implementations of this specification is:"

Jirka: add only "XML-based" to implementation

<fsasaki> ACTION: felix to go back to richard about new resolution for issue-79 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action06]

<trackbot> Created ACTION-393 - Go back to richard about new resolution for issue-79 [on Felix Sasaki - due 2013-01-30].

issue-80

Felix: we can just add links to example

<scribe> ACTION: felix to add links to examples for issue 80 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action07]

<trackbot> Created ACTION-394 - Add links to examples for issue 80 [on Felix Sasaki - due 2013-01-30].

issue-81

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0071.html

felix: related to issue-89

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0066.html

Felix: issue is not clear how HTML maps to ITS
... some HTML construct are explicitely mapped, other are not
... like terminology (dfn, dt, etc.)
... should an implementer of HTML/ITS process those constructs as term? or not?

<fsasaki> http://www.w3.org/TR/2008/NOTE-xml-i18n-bp-20080213/

<fsasaki> http://www.w3.org/TR/2008/NOTE-xml-i18n-bp-20080213/#relating-its-plus-xhtml

Felix: Possible solution is a mapping defined in bets practice
... like we did in ITS 1.0
... we did this only as a best practice
... e.g. we don't talk about dfn in ITS 1.0
... for issue 81 we would not define normative relation to term
... but provide mapping in best practices document

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0111.html

Felix: related issue is issue-97
... some HTML features are used but not declared as such, like 'translate'
... we should have something like "the ITS processor implementing Tranlsate MUST implement HTML5 translate attribute"

issue-82

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0067.html

Felix: if values are ok, no need to have a mapping

<fsasaki> http://lists.w3.org/Archives/Public/www-international/2013JanMar/0048.html

felix: something without mapping just pass through

<fsasaki> answer to the comment: "STEP 3-1-2-5-2. Else (if no mapping is found): Add the string (in its original cases) to the result string."

<scribe> ACTION: daveL to reply to Richard [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action13]

<trackbot> Created ACTION-401 - Reply to Richard [on David Lewis - due 2013-01-30].

case related comments

<fsasaki> https://www.w3.org/International/multilingualweb/lt/track/issues/102

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0116.html

Pablo: at first we used case-sensitive
... then we moved to insensitive
... we could compare directly
... but if document is encoded differently we may have entities
... and the string is different

<fsasaki> scribe: fsasaki

yves: by entity you mean "person"?

pablo: yes

<pnietoca> ... domainMapping="Economía (ECON), Leyes (Law)"/>

yves: but that gets resolved then you parse the documnt

pablo: see example above

yves: then you read the document the entity wil be converted into í
... if we just do case-sensitive we have a problem
... the reason why we want to have insensitive: to avoid duplicates
... because we know people don't regard casing for keywords anyway
... so in one case we say: case matters, in others we say they don't matter
... so one solution is: case always matters
... but what is the solution for HTML?

davidF: wouldn't be worried that you preserve case
... only if you fail to map

yves: only when you compare during the mapping you are uncertain
... problem is: many documents have keywords typed differently
... could also have a keyword saying "mapping or not"

felix: would that delay the problem
...

resolution: agree with first question in http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0116.html
... 2nd question becomes unnecessary

<scribe> scribe: Yves_

action yves to fix text and algo for domain case mapping

<trackbot> Created ACTION-402 - Fix text and algo for domain case mapping [on Yves Savourel - due 2013-01-30].

scribe Yves_

ISSUE-84

dF: dave split indic language issues into 3 topics
... first one is covered in issue-84

<fsasaki> reply from dave on issue-84 at http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0149.html

dF: answer is: yes transliterating is different but we didn't have enough use cases for a requirement
... that made it as a final data category

felix: so we are waiting for a reply now

ISSUE-86

felix: implementation committement

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0144.html

for several issues

scribe: for Ruby and Directionality
... basically we don't have experts and no volunteer to implement
... Ruby may be ported for XLIFF
... still not sure what is the aim: dropping ruby or not?
... also not sure when we can expect stability
... but we want to be feature complete very soon
... questions to the i18n are out, waiting for feedback

<fsasaki> yves: directionality is not really used in XLIFF

<fsasaki> .. implementers use control characters

<fsasaki> .. we tried really hard in XLIFF2

<fsasaki> .. we have a module for directionality in XLIFF2

<fsasaki> .. but the implementers would insert rather control characters than markup

dF: when we discussed directionality in Lyon, someone described how to do dir with inline markup

felix: .. for Ruby, I don't think anyone implemented the pointer for example

Arle: need to speak to Asian developers
... group is not representative
... for these issues

Felix: for Japanese there is a detailed document on layout
... and requirements in XML and HTML are pushed by this doc and issues not addressed in ITS2ument
... Our question is how can we deal with it?

Arle: maybe it can be defined later in a different namespace

Felix: maybe, but baiscally it's the same for ITS 2.
... lunh time now

<Arle> s/lunh/lunch

<Arle> s/lunh/lunch/

<Arle> Scribe: Arle

meeting schedule

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/EventSchedule

Felix: I thought of discussing the next meetings, but Pedro isn't here.
... See the wiki page. You will see that thanks to Tadej that we have a face-to-face in Bled in May.
... I just got an email from Pedro with some offers to host the face-to-face in Madrid, but all are beyond budget (€5000), because he would have to rent meeting space.
... We might need to think of an alternative to Madrid. One alternative is LocWorld in June in London.
... We could ask Microsoft if there is a London office we could use.

<fsasaki> s/Microsoft/xyz/

LocWorld is 12–14 June

David: 10 June is XLIFF; 11–12 June (?) is FEISGILTT

Felix: We will need technical discussions in June.

Yves: Whole week is booked for some people with the different events.

Felix: Week of 17th?
... Please check your calendars to see if that might work.
... 17–18 June is the suggestion.

Location: TBD in a cheap place.

Felix: Berlin would be free.

s/Location:/.. Location/

Dave: Dublin is an option.

<scribe> ACTION: Felix is to check availability of Berlin on 17–18 June. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action14]

<trackbot> Created ACTION-403 - Is to check availability of Berlin on 17–18 June. [on Felix Sasaki - due 2013-01-30].

<scribe> ACTION: daveL to check availability in Dublin for face-to-face meeting on 17–18 June. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action15]

<trackbot> Created ACTION-404 - Check availability in Dublin for face-to-face meeting on 17–18 June. [on David Lewis - due 2013-01-30].

Pedro: I am looking at various possibilities in Madrid still.

Felix: Would it be OK for you if we look at other cities to save costs?

Pedro: That is fine for me. Leave Madrid as an alternative.
... My latest option in Madrid comes to 3–3.5K€, if we have everyone stay at the same hotel.

Felix: We need to fix these dates as soon as possible because of Localization World so that travel can be arranged by everyone as appropriate.
... Dave and I will try to decide so people can make arrangements.

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/EventSchedule

Felix: We are also considering another face-to-face in September, around LRC conference.
... In Limerick.
... Dates would be 16–17 September (pending confirmation).
... Would 23–24 September be also good

<fsasaki> will come back to september meeting tomorrow

s/also good/also good?/

<fsasaki> 23-24 would be difficult for cocomore

Last workshop

Felix: Project ends in December. DoW shows we spend most efforts until September, so if the workshop is in December, mass may be difficult. Do we have a regular workshop, or some other kind of event?
... Any ideas of other options for final event?
... We can't drop it due to work package, which describes it as biggest workshop.

Pedro: What about colocation of the final workshop with another event?
... David: What about tcworld?

s/.. David:/David../

scribe: It is a big one. Might be good to connect there.

<scribe> ACTION: Felix to follow up with Christian on tekom as an option. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action16]

<trackbot> Created ACTION-405 - Follow up with Christian on tekom as an option. [on Felix Sasaki - due 2013-01-30].

Arle: Consider that colocating with a commercial event will likely have higher costs.

Felix: We can do another MLW workshop, or look at other options.

Yves: That is a lot of work.

Felix: Yes, and after September, we can't ask people for a lot of work.
... Also, September/October is probably too early for the next workshop after the one in March.
... What if we don't make a conference or go to one? Instead we have an event (possibly closed) to do demos to customers?

<Pedro> Pedro: Tekom, Wiesbaden 06Nov-08Nov2013

Felix: we can consider still in January. Let me and Dave know of any options that come to mind.

Dave: I can already confirm space would be available in Dublin in June.

posters

Felix: Our reviewers will most likely not be in Rome. So we need to make a presentation in Luxembourg. Posters would help show completion.

Pedro: What size should they be?

Felix: A0.

<scribe> ACTION: Arle to resize templates for posters from A1 to A0. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action17]

<trackbot> Created ACTION-406 - Resize templates for posters from A1 to A0. [on Arle Lommel - due 2013-01-30].

Issues

<daveL> scribe daveL

issue-88

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0070.html

<daveL> felix: this is just editorial in the directionality section

<scribe> Scribe: Arle

<fsasaki> s/topic: Issues//

David: I don't know the difference between the HTML elements here.

<scribe> ACTION: Felix to check for clarification on Issue-88 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action18]

<trackbot> Created ACTION-407 - Check for clarification on Issue-88 [on Felix Sasaki - due 2013-01-30].

issue-92

<fsasaki> original mail at http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0076.html

Yves: This is a note from Richard asking why information is in a note, which is not normative.
... Can a note be normative? I believe they can be if they are in a normative section. I believe we have MUSTS in notes.

Felix: I think that is a mistake.

<scribe> ACTION: Felix to ensure that there is no MUST in any notes. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action19]

<trackbot> Created ACTION-408 - Ensure that there is no MUST in any notes. [on Felix Sasaki - due 2013-01-30].

Yves: idValue global has one.

<fsasaki> relation to issue-103 - clarify the algorithm

Yves: One explanation + bullet explaining that empty = no locale and * = all locales. Then we can eliminate the note.

Felix: Solution is to have three bullets explaining the cases, and delete note. Resolves issue-92 and issue-103.
... Yves, do you use extended filtering?

Yves: Yes. We do. We need to check with Shaun, but I believe this is the algorithm for extended filtering.

Felix: We need to express the approach described in BCP47 and that it will work for everyone implementing this. Tilde should check.
... Ankit and Marcis, should we return to this, or can we assume that if we don't hear otherwise, it’s OK?

<scribe> ACTION: Yves to follow up with Richard and Norbert on issue-92 and issue-103. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action20]

<trackbot> Created ACTION-409 - Follow up with Richard and Norbert on issue-92 and issue-103. [on Yves Savourel - due 2013-01-30].

Issue-93

Jirka: Proposed resolution is to use what was proposed by original commenter.

<scribe> ACTION: Jirka to write to Henry on issue-93 and make the change in the text. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action21]

<trackbot> Created ACTION-410 - Write to Henry on issue-93 and make the change in the text. [on Jirka Kosek - due 2013-01-30].

Issue-94

Felix: I think Jirka has a proposed resolution.

Jirka: I've sent replies to Henry, but not heard back. I think we should resolve this issue in a different way. See link at end of issue.
... HTML has different rules for processing white space and decimal numbers. There is different precision between XML and HTML.
... The easiest resolution is to use the double data type in XML for ITS. It will align XLM and HTML. Double is implemented in almost all programming languages. So we move all data types to double and deal with the differences in leading and trailing whitespace between the two.

Felix: This impacts localization quality, MT confidence, and localization quality rating.
... Is this OK for all implementers?

Jirka: Only difference is that double has lower precision than decimal. And you can use exponential notation.

Felix: Also disambigConfidence and term confidence.

<scribe> ACTION: Jirka to change localization quality, localization rating, mt confidence, term confidence, and disambig confidence to use double rather than decimal and respond to Henry (Issue-94) [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action22]

<trackbot> Created ACTION-411 - Change localization quality, localization rating, mt confidence, term confidence, and disambig confidence to use double rather than decimal and respond to Henry (Issue-94) [on Jirka Kosek - due 2013-01-30].

issue-95

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0093.html

Felix: We should reject this. The proposal itself said that translatable is different than localizable (e.g., in formatting numbers and images).
... Discussion was between Norbert, Felix, Des, and Phil.
... I think addressing this would take too much time at this point.

<fsasaki> another point for Dave here http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0147.html

Dave: It really is out of scope for ITS.
... Translators will deal with this on their own anyway.

Felix: Norbert asked if we could use ITS for localizing CLDR? I don't see that as a real use case.

<scribe> ACTION: Felix to let Norbert know that action-95 is out of scope. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action23]

<trackbot> Created ACTION-412 - Let Norbert know that action-95 is out of scope. [on Felix Sasaki - due 2013-01-30].

issue-98

<fsasaki> s/issue-98/issue-98 and issue-99

Milan: related to issue-99. I found that there is no way to do this. It is mentioned only for global approach to selectors and what is allowed. Chapter 1.1 should state that the local approach can be applied only to the content of the current element and any inherited nodes, per 8.1
... For issue-99, when using selectors in ITS, how do you select attributes? Information is there, but the definition of node differs between XML and HTML, leading to confusion. I see Yves’ suggestion to remove CSS as a selector type since they can point only to elements, but I would keep it and add a note that we can only point to elements, not attributes.

David: I think it makes sense to keep CSS.

Felix: We don't have any implementers using selectors.

Yves: Shaun is, as a prototype.

Felix: I never got it to work.

Yves: Norbert says for HTML people selectors may be important.
... But with no implementations, it won't happen. It's marked as endangered.

Felix: We can drop "at risk" bits.
... I agree with Milan's solution, but we might drop them anyway.

Jirka: suggested a path to get implementation.

Felix: It would be nice. Right now we have two paths, doing testing only for XPath, but not for CSS.

Jirka: Do we need tests, since they just select nodes?

Felix: Maybe the test suite or elsewhere, would we have examples making use of CSS.
... If we don't have testing, W3C management may not like us saying "you can do it on your own but we haven't done it."

Jirka: We need at least one selection mechanism. Testing is to verify interoperability.

Felix: We need to have at least one example for standardization and users about how to use it. We have no CSS examples.

Jirka: Let's have some examples, parallel to XPath examples.

Felix: Can you link to libraries to convert between CSS and XPath selectors?
... Are there non-browser conversions?

<scribe> ACTION: Jirka to find data on CSS and XPath selectors conversion libraries and keep CSS selectors in the spec. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action24]

<trackbot> Created ACTION-413 - Find data on CSS and XPath selectors conversion libraries and keep CSS selectors in the spec. [on Jirka Kosek - due 2013-01-30].

issue-100

<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0126.html

Felix: Yves proposed a resolution.

<scribe> ACTION: Felix to make edit for issue-100 and get back to Norbert. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action25]

<trackbot> Created ACTION-414 - Make edit for issue-100 and get back to Norbert. [on Felix Sasaki - due 2013-01-30].

issue-104

<scribe> ACTION: Felix to update unicode reference for issue-104 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action26]

<trackbot> Created ACTION-415 - Update unicode reference for issue-104 [on Felix Sasaki - due 2013-01-30].

issue-106 and issue-107

<fsasaki> 106 see http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0121.html

<fsasaki> s/106/107/

Karl: Norbert asked some questions and we weren't sure how to resolve them. It isn't up to the spec. The implementation must support UTF-8, but that is up to the implementer. It is best practice, especially for storage size. But we don't think it has to be mandatory for all implementations.

<fsasaki> 106 see http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0120.html

Karl: Other question was how to handle encoding when the implementation doesn't support it. Again, this is not up to the spec. We can define best practice, but it doesn't need to be stated in the spec.

Stephan: Perhaps we have an explanation about what storage size is used for. The question is about when it is used to markup text in the source language. It is informational, but not up to the spec to tell us what to do if a tool doesn't support an encoding or if user text cannot be represented in a given encoding.

Karl: We should add a sentence to storage size, per the note on the issue-107.

Felix: on issue-106 and issue-107 we do nothing, just let Norbert know the rationale.

<scribe> ACTION: Karl to propose solution to Norbert and then Felix can add to spec. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action27]

<trackbot> Created ACTION-416 - Propose solution to Norbert and then Felix can add to spec. [on Karl Fritsche - due 2013-01-30].

Felix: When we go back to Norbert, talk about what we did in the group to show there is consensus.

issue-108 and issue-109

Felix: Both relate to Indic requirements.

Dave: They make a point that there is dependency on context (e.g., part of speech) that influences how you translate things. They want PoS in localizationNote and provided an annex of possible annotations.
... Adding a data type specifically for this would be a big change. You see companies when they want to add their own metadata use localizationNote with name:value pairs. It could be best practice outside the spec.

<daveL> http://lists.w3.org/Archives/Public/public-multilingualweb-lt-comments/2013Jan/0154.html

<fsasaki> reply from Dave on locNote its2 req , see above mail

Dave: I pointed them to other relevant resources, like NIF.

Arle: This would be too complex for us to solve this problem. Anything that works for Europe may fall apart elsewhere.
... I don't think we could solve this in a reasonable time frame without too much controversy.

Tadej: they have PoS taggers in MT already, but it is specialized. This would be scope creep.

Marcis: Once you add PoS, you have to add syntax, etc.…

Dave: Do humans need PoS tagging? I don't know.

Marcis: Wouldn't this be duplicating existing work in text analysis.

<scribe> ACTION: DaveL to go back to Somnath on issue-108 to explain why we won't address it. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action28]

<trackbot> Created ACTION-417 - Go back to Somnath on issue-108 to explain why we won't address it. [on David Lewis - due 2013-01-30].

Dave: issue-109 falls out of my expertise. It deals with nested output from NER.

Tadej: I didn't quite follow the requirements. It seems they want to show that parts of entities may be entities. I don't know if they need this or are showing what they might do with this.
... Regardless of this, the comment that hierarchy is needed.

Dave: We can't do this.

Tadej: Overriding makes that the case, but if we allowed multiple values, we could.

Dave: But you need to show that the different parts are bound together.

Tadej: If you allow multiple values (e.g., something can belong to two entities), then the scope can be ambiguous.

Marcis: But there should be no ambiguous overlaps in a hierarchy.

Stephan: When would you actually use the knowledge that you have nested named entities?

Tadej: Can we make the restriction that entities are contiguous?

Dave: That would be reasonable.
... The solution isn't straight-foward. This would be a new feature. I think we should respond in that way.

s/Dave: The solution/.. The solution/

Discussion about whether hierarchy is needed and produced.

Dave: You could also point to a NIF record with that structure in it.

Tadej: If several disambiguationRefs address something, we can't tell which one produced what.
... If a single node can have multiple values it makes tracking hard. We use stand-off for this.
... This multiple granularity might break things.

<scribe> ACTION: Dave to respond to Somnath on issue-109 to explain we are looking at it to make recommendations. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action29]

<trackbot> Error finding 'Dave'. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.

<scribe> ACTION: DaveL to respond to Somnath on issue-109 to explain we are looking at it to make recommendations. [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action30]

<trackbot> Created ACTION-418 - Respond to Somnath on issue-109 to explain we are looking at it to make recommendations. [on David Lewis - due 2013-01-30].

locale filtering question

<fsasaki> marcis: in content is "de"

<fsasaki> .. in the localeFilter it would be de-de

<fsasaki> felix: not matched

test suite check

Felix: We don't have a lot of coverage (38%) and most of that is thanks to Yves and Fredryk (ENLASO).
... At the end of January we have the deadline to run all test cases. Is that deadline (next week) realistic? We have some changes, but others are stable.

Leroy: The files will remain the same, with changes after the 21st.

Karl: our cases are theoretically all working, but we have some issues with sorting of attributes, which we don't do. That's the only reason we aren't complete.
... In the input attributes are source and alt. We output them in that order, but the output sorts them.

Leroy: I can run my sorting function on output for you.

Stephan: Actually, it is backward, the source is in order, the output isn't.

Yves: Many engines do not care about order. You have to handle sorting yourselves.

Karl: It's not a big change and then we are done. I will make the change myself.

Ankit: We have a few small snags.

Linguaserve: (Some issues. ???)

Thomas: We are working on our implementations, should be ready next week.

David: Connection between Moravia and UL tests…

Felix: David, I know you use Okapi wrapper. When that is integrated in the workflow, you can run the same tests as Okapi. So now you run six cases, but you could run more then.

RFC statements

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements

Felix: Much is covered by the schema.
... #25 talks about the content of the annotatorsRef attribute. Currently the data type is text. There is a need for test case with a file with a non-allowed identifier and the parser says it is wrong. That would test it, even though it does not produce specified output.
... David, could you make a test case and get the implementers to run it?
... See example below:

<fsasaki> annotatorsRef="mt-confidence|tool1"

<fsasaki> annotatorsRef="mtconfidence|tool1"

Felix: Second line should throw an error.

Yves: Do we have standard output for the errors?

Felix: No. This will require human verification.
... We can address issues here until October.
... After XML Prague would be fine.

Jirka: We can do this using Schematron with regex.

Karl: There are similar cases in the docs to do negative tests.

Jirka: It's already there, but you have to look at the Schematron, not the XSD.
... Doing as much as possible in Schematron.

Felix: What about #39, #35, #41?
... If not checked by Schematron, please add later.

<fsasaki> ACTION: jirka to make schematron tests described at http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements#Purpose [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action31]

<trackbot> Created ACTION-419 - Make schematron tests described at http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements#Purpose [on Jirka Kosek - due 2013-01-30].

Felix: #31, if values have spaces, must be delimited with quotation marks. Need a test case?

Yves: It's already covered by the test cases, which fail if the output isn't formatted properly.

Felix: #36. Overriding means these won't be combined anyway. Maybe make an action to delete the sentence in 8.11.2?

Action-420

Refers Issue-111

<fsasaki> ACTION: felix to make edit for issue-111 [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action32]

<trackbot> Created ACTION-420 - Make edit for issue-111 [on Felix Sasaki - due 2013-01-30].

Felix: #36 is dropped.

<fsasaki> " If the type of the issue is set to uncategorized, a comment MUST be specified as well." - can be checked, an error if no comment is avaiable

Felix: Maybe we put the other MUST statement (about mapping internal types to issue type values) as its own test type. To catch the error, you must be able to parse the category.
... You need to understand the values and different types or markup. It is on top of the normal test suite functionality.

Yves: We don't need the MUST there. The value column covers the same thing.

Discussion about where to test.

test suite

<fsasaki> s/topic: test suite//

<fsasaki> "The set of characters that are allowed is specified using a regular expression. That is, each character in the selected content MUST be included in the set specified by the regular expression."

<fsasaki> this is not a test for the processor, but for the consuming application

<fsasaki> for IANA charset names see http://www.iana.org/assignments/character-sets/character-sets.xml

<fsasaki> we point to the IANA list, that's it

<fsasaki> relevant for this MUST statement: "A storageEncoding attribute. It contains the name of the character set encoding used to calculate the number of bytes of the selected text. The name MUST be one of the names or aliases listed in the IANA Character Sets registry . The default value is UTF-8."

Felix: For many quality issue type items, change MUST/MUST NOT to must/must not.
... Numbers 45–48

<fsasaki> "See entries 45-48 at http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements#Purpose these statements are not verifable. Proposal is to set MUST and MUST NOT to lower case to make clear that the text is just guidance."

<fsasaki> for 45 " The values a tool implementing the data category produces for the attribute MUST match one of the values provided in this table and MUST be semantically accurate.": re-formulate this :

<fsasaki> drop "MUST be semantically accurate".

"If a tool can map its internal values to these types it MUST do so and MUST NOT use the value other, which is reserved strictly for values that cannot be mapped to these values." -> "Note that the other category is reserved for cases where a tool-specific category cannot be mapped…"

<fsasaki> ACTION: arle to work on statements 45-48 at http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements#Purpose , see prague f2f minutes [recorded in http://www.w3.org/2013/01/23-mlw-lt-minutes.html#action33]

<trackbot> Created ACTION-421 - Work on statements 45-48 at http://www.w3.org/International/multilingualweb/lt/wiki/Ownership_of_rfc2119_statements#Purpose , see prague f2f minutes [on Arle Lommel - due 2013-01-30].

Yves pointed out that the values should be done by class, not on an individual error basis independent of classes.

#48. If a system has an "miscellaneous" or "other" category, it MUST be mapped to this value even if the specific instance of the issue might be mapped to another category -> append note on semantic accuracy here.

requirements doc

<fsasaki> multi-engine domain scenario + multi engine domain scenario

<fsasaki> issue-95 and issue-75 would be covered by this

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#autoLanguageProcessingRule

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#Process

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#formatType

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#genre

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#purpose

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#translatorQualification

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#register

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#contentLicensingTerms

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#author

<fsasaki> (covered by dc.terms

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#confidentiality

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#context

<fsasaki> http://www.w3.org/International/multilingualweb/lt/wiki/Requirements#languageResource

<swalter> for 45: Note that the other category is reserved... -> Note that the "other" category is reserved to cases where a tool-specific category cannot be mapped to any of the first categories in a semantically accurate manner.

- DRAFT -

MLW-LT f2f

23 Jan 2013

Attendees

Contents

roll call

http://tinyurl.com/its20-comments-handling

issue-67

issue-69

issue-70

issue-71

ISSUE-72 NIF comment

issue-68

issue-75

issue-73

issue-74

issue-72

issue-76

issue-77

issue-76 again

issue-78

issue-79

issue-80

issue-81

issue-82

case related comments

ISSUE-84

ISSUE-86

meeting schedule

Last workshop

posters

Issues

issue-88

issue-92

Issue-93

Issue-94

issue-95

issue-98

issue-100

issue-104

issue-106 and issue-107

issue-108 and issue-109

locale filtering question

test suite check

RFC statements

test suite

requirements doc

Summary of Action Items