See also: IRC log
<fsasaki> ACTION: felix to follow-up on discussion about regular expression for allowed characters, see http://www.w3.org/2012/10/31-mlw-minutes.html#item02 [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action01]
<trackbot> Created ACTION-261 - Follow-up on discussion about regular expression for allowed characters, see http://www.w3.org/2012/10/31-mlw-minutes.html#item02 [on Felix Sasaki - due 2012-11-08].
Agenda accepted
<fsasaki> people doing self-intro
<fsasaki> http://www.w3.org/International/multilingualweb/lt/drafts/its20/its20.html#domain-implementation
<Yves_> changes for domain are: removing quotes and duplicates
<fsasaki> Need for a common representation of the ITS data categories in XLIFF, see XLIFF_Mapping
felix: domain issue is "fine"
<Yves_> ITS to XLIFF working draft mapping: http://www.w3.org/International/multilingualweb/lt/wiki/XLIFF_Mapping
Implementors are waiting for the XLIFF mapping
XLIFF
current direction is that pointers are not be needed in XLIFF
<fsasaki> "mrk", "sm" and "em" elements in XLIFF. "mrk" and "sm" would be extended
<fsasaki> XML Schema subset of regex for allowed characters
<Yves_> the issue is that not all programming languages support all XML schema regex pattern
<fsasaki> http://www.w3.org/2012/10/31-mlw-minutes.html#item02
we keep the draft as it is now and postpone the topic of regular expression in allowed characters
<fsasaki> Allowed Characters regular expression for not allowing HTML tags in content nodes where only plain text content is allowed
<fsasaki> ACTION: pedro to write note about allowed characters issue with help from Mauricio and karl - due 8. November [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action02]
<trackbot> Created ACTION-262 - write note about allowed characters issue with help from Mauricio and karl [on Pedro Luis Díez Orzas - due 2012-11-08].
<fsasaki> "Need to add all the document HMTL tags (wrap the content with html, head, body tags) so we can add a link to a global rules XML "
<fsasaki> http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Nov/0005.html
<scribe> ACTION: implying some its:rules in html tags [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action03]
<trackbot> Sorry, couldn't find implying. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.
<fsasaki> http://www.w3.org/TR/xml-i18n-bp/#relating-its-plus-xhtml
<fsasaki> 1) input 18_8.xml
<fsasaki> 2) XML parsing of HTML fragment e.g. via validator.nu library in 18_8.xml
<fsasaki> output of step 2): DOM or XML serizalization
<fsasaki> 3) normal ITS processing
<fsasaki> 0) input is "some HTML data"
<fsasaki> next step: step 2)
<fsasaki> next step: step 3)
Coffee break now
<Yves_> scribe: Yves_
mauricio: one possibility for solving this could be to changing to XHTML the HTML content
dF: that would allow local ITS markup
Felix: so no change for the data
category
... maybe some note on best practices?
dF: one problem Karl noted is the validator inside Drupal
Jirka: I can try to look at it and see it could be fixed
Felix: but do we need guidance for CMS in general?
dF: maybe in a non-normative section
Moritz: yes a best practice
<fsasaki> ACTION: david to summarize the options and the recommendations related to HTML parsing workflow in the CMS, see discussion at http://www.w3.org/2012/11/01-mlw-lt-irc#T10-13-41 [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action04]
<trackbot> Sorry, ambiguous username (more than one match) - david
<trackbot> Try using a different identifier, such as family name or username (eg. dlewis6, dfilip)
<fsasaki> ACTION: dfilip to summarize the options and the recommendations related to HTML parsing workflow in the CMS, see discussion at http://www.w3.org/2012/11/01-mlw-lt-irc#T10-13-41 [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action05]
<trackbot> Created ACTION-263 - Summarize the options and the recommendations related to HTML parsing workflow in the CMS, see discussion at http://www.w3.org/2012/11/01-mlw-lt-irc#T10-13-41 [on David Filip - due 2012-11-08].
Felix: so maybe a best practice for CMS
<fsasaki> "Troubles with namespaces in HTML5. "
<pnietoca> xmlns:h="http://www.w3.org/1999/xhtml"
Pedro: Jirka noted we need the namespace definition
<pnietoca> http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Oct/0226.html
Jirka: the namespace declaration
is missing
... probably an implementation issue
... the node in the DOM should list the namespaces
... I can help if needed
Richard Ishida comes to discuss logistisc/topics with HTML5 WG
scribe: and I18N WG
<fsasaki> tomorrow 11 a.m. meeting with i18n wg
back to implementation issues
scribe: HTML5 namespace issue resolved it seems
<fsasaki> "Need to come to an agreement to map domain values to be consistent for both Lucy and DCU's MT Systems. "
Felix: important for the showcase that the two system communicate
<fsasaki> ACTION: thomas to follow up on domain list topic with DCU [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action06]
<trackbot> Created ACTION-264 - Follow up on domain list topic with DCU [on Thomas Rüdesheim - due 2012-11-08].
<fsasaki> "Problems to use global rules with the provenance and quality metadata since ATLAS PW1 cannot place files on the client server. "
Pedro: our system doesn't allow
to create rules file on the client
... rules file then points to our server, but that's not the
right way to do it
... also will we drop global rules?
David: important question for
some data categories like provenance, etc.
... some idea is to exclude some pointers
... but some use cases show some non-pointer global rules are
useful
... other issue is how to address attributes
... transltable attributes is not best practice,
... so maybe it's ok to drop global rules despite those use
cases
Felix: what about the quality issue
dF: well some attributes are not
going away, like in HTML5
... no way to markup things like alt or title
Felix: two types of implementers:
some using XLIFF extration, other work directly with original
data
... for XLIFF case this is not an issue
... for original data this is a problem
David: we should be consistent,
if an attribute can be translated, it should be able to get
other data categories too
... an approach maybe could be to use a local attribute?
Felix: two topics one is where to
put the global rules, the other is do we need global
rules
... with local, you can't address attributes
Tadej: so global rules in
script
... ok with that solution
... also concern about a rule should not have pointer and
values at the same time
... basically not doing proxy stand-off annotation
Felix: yes, should we drop this
statement?
... one could have combination of both in some cases
Tadej: this would cause clashes
in some cases
... not sure if this affect stand-off
Felix: stand-off is
different
... the URI points to a piece of info, so it still "add"
information
Tadej: so goal to to keep all info in one place rather than scatter it around
Felix: If we keep the constraint what do we do with cases like Quality Issue
David: is XLIFF the only use case for pointer in Quality issue
<fsasaki> yves: pointer for the reference attrbute to standoff markup is needed
<fsasaki> .. at least in the XLIFF case
<fsasaki> felix: would that be resolved by a mapping table?
<fsasaki> yves: would not make the stuff processable by an ITS processor
<fsasaki> yves: need for pointer would be just for reference attribute, not the other pointers
David: if XLIFF is the only use case, should we make it a special case?
<fsasaki> yves: can imagine an "Its only" processor to deal with quality issues
<fsasaki> .. they can have special information for ITS
<fsasaki> .. but big question is: do we want ot allow that to happen
<fsasaki> .. that is have other formats to work without XLIFF markup
<fsasaki> .. if other formats don't allow the ITS native mapping attribute, what to do?
<fsasaki> .. if XLIFF mrk allows extension, the problem goes away
Felix: difficult to discuss because we depend on XLIFF decission
<fsasaki> yves: we need to wait for next week XLIFF meeting, if is not resolved, we need to find a different way
<fsasaki> ACTION: daveL to react to decision on extensibility for mrk in XLIFF and the result for our "pointer" attributes [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action07]
<trackbot> Created ACTION-265 - React to decision on extensibility for mrk in XLIFF and the result for our "pointer" attributes [on David Lewis - due 2012-11-08].
Felix: Example 54 and 55
... RDFa
... .. ITS generic processor can then work with RDFa
... Karl provided feedback
... is this a use case for Pointer
Tadej: the example as some small issue
<fsasaki> "typeof=http:/nerd.eurecom.fr/ontology#Place": only single slash
Tadej: single slash missing and miss-named attribute
<tadej> Example 55: "entityTypeRefPointer" -> "disambiguationClassRefPointer"
<tadej> "disambigClassRefPointer"
Felix: is that the only way to
consume RFDa if you don't know about it?
... it seems so
Tadej: yes
... main difference with Example 52
... it's much less stable, more prone to error
... using the pointer is more stable
... but see why 52 would not be recommended
... better for consistency using local would be
better
<fsasaki> ACTION: felix to edit disambig example [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action08]
<trackbot> Created ACTION-266 - Edit disambig example [on Felix Sasaki - due 2012-11-08].
Felix: Currently
pointer/non-pointer can be mixed in Disambiguation
... not good
... you should do either one
Tadej: agreed
<fsasaki> ACTION: felix to try to simplify disambiguation globally [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action09]
<trackbot> Created ACTION-267 - Try to simplify disambiguation globally [on Felix Sasaki - due 2012-11-08].
<fsasaki> "Language Information: Use-Case? When is xml:lang or lang not enough or can't be used? "
karl: we have xml:lang or
lang
... when using HTML we should use lang
... maybe the content of those attributes should be explicitely
defined
Felix: in section 6.7.1
... we say that BCP47 is the value
Dave: it's there for XML
vaocabulary where xml:lang is not used
... no need for pointer in HTML5
Pedro: another case is to keep info about what was the original language
<fsasaki> http://www.w3.org/TR/xml-i18n-bp/
<fsasaki> Example 19: Declaring language information with a non-standard mechanism
Pedro: but that's not for this.
Felix: yes, it's for XML
Pablo: was not able to come up with use case in HTML
Dave: so all three xml:lang, lang and pointer are relevant
<fsasaki> yves: various legacy content has other attributes for language information, that is a use case for language Information
<fsasaki> .. all those cases use bcp 47 as a value, so with the pointer attribute we always get the same value
dF: what about the BCP47 extensions
Felix: ITS does not do validation
of content
... could have a note about issue related to extensions
Pedro: maybe ITS 3.0 could have a way to indicate the original language
Felix: could be done in provenance maybe
Moritz: so can we drop langInfo for HTML5?
<fsasaki> "[Ed. note: Add something about HTML5 lang]"
Felix: there is an editor note about it
<dF> http://tools.ietf.org/html/rfc6497
<scribe> .. Done with the implementation issues
Felix: could start with test suite
Jirka: we need to start to work
on schema
... several data categories are still not stable
... we need that for the publication
Felix: for disambiguation MTconfidance it's about pointer and tool reference
for LQ Precis it's about implementation commitements
dF: name may prevent adoption
Felix: another detail is the
question about global rules
... let's say cut off date in 3 weeks
... including the details with global rules and pointers
Jirka: maybe too short, need 2
weeks at least
... could do it in parallel
felix: Friday 23rd would be the
cut off day
... then end of nov we have stable content
... then we have two more weeks for schema and tests
... not sure how realisitic it is in december
Leroy: should be ok for tests
Felix: we want to get to LC
... and we have a January f2f where we want to address
comments
... if we send this at start of December people would have time
for commentd
... should we have a call in week of 26th for test/schema
Jirka: not sure,depends on time I need to work on schema
<fsasaki> ACTION: dom to make sure that schedule for test suite and schema update discussed at http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken into account [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action10]
<trackbot> Sorry, couldn't find dom. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.
<fsasaki> ACTION: DomJones to make sure that schedule for test suite and schema update discussed at http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken into account [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action11]
<trackbot> Sorry, couldn't find DomJones. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.
<fsasaki> ACTION: Dominic to make sure that schedule for test suite and schema update discussed at http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken into account [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action12]
<trackbot> Created ACTION-268 - Make sure that schedule for test suite and schema update discussed at http://www.w3.org/2012/11/01-mlw-lt-irc#T11-27-30 is taken into account [on Dominic Jones - due 2012-11-08].
Pedro: LQ precis could be LQ metrics
dF: could bs LQ score
felix: lunch now. back in one hours
<daveL> scribe: daveL
DomJones: breif overview of test suite
<DomJones> https://docs.google.com/spreadsheet/ccc?key=0AgIk0-aoSKOadG5HQmJDT2EybWVvVC1VbnF5alN2S3c#gid=0
DomJones: looking at these tables there are some features that don't have two implementors
<scribe> ACTION: DomJones to ask Phil to confirm whether or not he will implement provenance [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action13]
<trackbot> Sorry, couldn't find DomJones. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.
Yves: not committing yet to provenance as it is not stable
Dom: Yves just confirmed he will support disambiguation
Tadej: will implement text analysis annotation if it is stable
dF: seems to be too unstable, needs to be resolve in specification section
Dom: quality issue is partially covered by two implementors, but different useages of using OKAPI is acceptable.
felix: quality has stand-off in
script as well as in-line option. Do we want both, as these are
not reflected in the test suite currently
... this choice has a lot of knock on for test suite
Yves: prefer the stand off in script over inline
<scribe> ACTION: Dave to check with Phil what his preference was between quality issue locally with inline and script-based stand off [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action14]
<trackbot> Sorry, couldn't find Dave. You can review and register nicknames at <http://www.w3.org/International/multilingualweb/lt/track/users>.
<scribe> ACTION: daveL to check with Phil what his preference was between quality issue locally with inline and script-based stand off [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action15]
<trackbot> Created ACTION-269 - Check with Phil what his preference was between quality issue locally with inline and script-based stand off [on David Lewis - due 2012-11-08].
dF: the script based appraoch may still recieve negative reaction from the HMTL WG
Jirka: HTML WG will say to do
this using microdata, but this doesn't work in this case
... HTML parsing consdere XML is script like it was CDATA
... but best way is the link to an external file
... or use XHMLT rather than HTML - which may not suite
everyone
<fsasaki> its-rules and its-standoff references?
Yves: the refernece for standoff mark-up could be an external file
felix: so in html would we have a separate link relation, in additon to its-rules, for such mark-up
<fsasaki> felix: not sure if above is needed, just *one* solution
df: prefer a span-based inline solution, vs a standoff in a script
jirka: if using the inline version, rdfa or microdata would be better, but HTML WG has stabalised this choice enought for us to make a decision
felix: but to summarise, for
cocomore/linguaserve, the script is better since propagation
back to client can be controlled more clearly
... so there is slight preference for script solution in the
room
Jirka: agrees it not a nice solution but the best we can manage given the problem of coexisitng between XML and HTML in general
felix: is there a volunteer to
explore the use of external file for standoff
... back to test suite
DomJones: now have two
implementers for quallity issues
... quality Precis, has only partial coverage from UL and
vistaTEC
df: only interested if someone produces it
yves: there was interest from Des also in this
df: this seems still quite unstable, for example how valid is the value of the score without the (optional) tool info
<fsasaki> ACTION: felix to ask phil and des and arle about need and implementation committment for localization precis during next call [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action16]
<trackbot> Created ACTION-270 - Ask phil and des and arle about need and implementation committment for localization precis during next call [on Felix Sasaki - due 2012-11-08].
dF: so this need more consideration especially about interpretation of score
felix: highlight the feature at risk in a draft, which allows us to cleanly remove after feature freeze - but not to change the feature
DomJones: mtconfidence and
alloewed text
... have also been now supported with imeplmentation
... asks what language people are using
... repsonses from room, Java, Javascript, php
... so tcd will try and provide some test sample code
...
... aim to have an online tutoril to help people with coding
the test suite
... suggest week 3rd December, suggest on Tuesday 4th
december
yves: need to avoid XLIFF call 4pm GMT
DomJones: so aim for 2.30pm GMT, 3.30 central european time
<fsasaki> http://www.timeanddate.com/worldclock/fixedtime.html?iso=20121204T14
DomJones: we will record this anyway for epople who miss it
<fsasaki> above wordclock link is the time of the webinar for the test suite
DomJones: now plan to switch to github for all the files and use google docs as index sheet and recording
Leroy: tst output will change to
be tab delmited and order the output alphabetically as suggsted
by Yves
... also use its as prefix in all cases
DomJones: when will files be frozen?
felix: after feature freeze, changes are unlikely, though external comment may require some changes.
<fsasaki> its-term="yes" vs. its:term="yes"
<fsasaki> http://www.w3.org/2012/09/mlw-lt-charter.html
felix: according to charter, testing can continue until october, but we should aim to be completed by March
<fsasaki> its-term="yes" vs. its:term="yes"
felix: clarification on prefixes in all cases is 'its:'
yves: could just drop prefix altogether
felix, leroy: agreed
DomJones: we will in new year work on making testing more accessible to implementors outside of the WG, as part of the general promotion of the of the spec and to encourage its uptake
felix: will there be some documentation on this
Leroy: yes there will be some in the github and on the web page
felix: will the test parser we made available
DomJone: yes will will release that later targetted at non working group members
mhellwig: explains demo
... two use cases
... one where client have few locaslaition staff, so they needs
tools to add its marks
... second is where staff can enhance content with ITS, but can
change the content directly - this is actually higher priority
for client
<fsasaki> scribe: fsasaki
karl: demo - editing
content
... in the body I can add the content and add metadata
... demo shows clicking on content, then adding a localization
note
... also specifying concepts, e.g. disambiguation target
... can also mark content that should not be translated or only
for certain languages
... that is lcoale filter metadata
... now about global metadata
... domain, revision agent, translation agent
... can add translate rules
... one field for the translate selector
... we think about helping the users by helping with creating
selectors
... another example of outputing the content as only
HTML(5)
... this is one option of output, the other is the XML file
that we have seen from Linguaserve before
... now there is an example with checking the
annotations:
... one can see the page and click on data category buttons,
then you can see the metadata available
... we now press a button and send the data to linguaserve
(switching machines, now working with Linguaserve machine)
Mauricio: now demo environment,
internal workflow interface
... here we have received the file from Cocomore
... it has information to transform the drupal XML file to CAT
tool oriented files
... now executed the preparational step
... now simulating the translation of the file
... assuming that the file has been translated now
... setting simulation translation marks, will explain these
later
... the file is now ready to be downloaded by Cocomore
pedro: we don't go into their
servers - we act as a server
... we don't go to the client's server
karl: now I am starting a manual
cron job
... to check if there are new translations
... the cron job checks regularly if there are new
translations
... the user doesn't need to do anything
... issue with cron job is that drupal has cron jobs
internally, so it takes some time until drupal gets to our cron
job
(karl's machine loading)
karl: translation received
... status changed from "in progress" to "needs review"
... now looking at the spanish node
... the language mgmt tab shows that all metadata is still
here
... there is also additional metadata: revision agent and
translation agent
(demo continues with Mauricio's machine)
mauricio: now showing what
happens with the XML file on our side
... current page is linked from our use case description in the
wiki
... now looking at the XML file we saw before -
21_11_orig.xml
... that we saw this morning
... now pre-production step
... in the log there is the XPath of the nodes, will be used
for the test suite
... currently there is only "translate" metadata test suite
output in here, others will come later
... now cat-tool oriented XML file
... in this file you have various information pieces, e.g.
domain information that the translator will see
... inside the translatable content there is some marks about
parts that are not translatable
... now showing a really translated file
... now using this for the post-processing step
... now you see the translators file in the original format of
the client
... what has changed: e.g. the "ready to process" in the
readyness has changed
... orignially it had four stages: hTranslate, ...
... now the state is: publish
... that means it can be published on the client side
... the time stamp related to readyness also has changed
... also we have now translation provenance information
... local xml:lang attribute has changed from "de" to
"es"
... next showing what will happen with the XML that we used to
translate
... content part that was marked as "translate=no" is blocked
so that the translators cannot change it
... also storage size is part of the file
... that is a file used in the transit translation tool
... once the engine does post-processing, the storarge size
will be applied
pedro: that is a simple way to
allow the translator to control the number of characters that
are possible
... this is just one cat tool, but that can be done with other
cat toosl as well
... expansion of global rules has been done before, and will be
done in post-processing
<mdelolmo> Provenance XPath selector value: Does the semantic combination between the Agent Provenance and Translate data category rules validate the regular expression for Provenance (//item)?
<mdelolmo> <its:transProvRule selector="//item" transOrg="Linguaserve" transPerson="LSTranslator01" transRevOrg="Linguaserve" transRevPerson="LSRevisor02"/>
<mdelolmo> <item id="18-body-0-format" its:allowedCharacters="." its:translate="no"><![CDATA[full_html]]></item>
discussion about interrelation between data categories
yves: you can imagine scenarios
where tools would implement only one data category
... would be better to have an XPath expression that only
selects the relevant nodes for provenance
... but that is probably marginal
<scribe> scribe: DomJOnes
Pablo: Global use of domain, second domain, mixture of translate / domain.
… sample is allways the same with different use of tags.
… no ITS errors identified. Click translate, same page presented but translated.
Pedro: Engine is supported by
Lucy SW. Thomas absent, behind this is Lucy S/W.
... Showing translate local
… examples in source text. Tags (translate yes/no) are set randomly.
… showing example of translation based on ITS tag (translate yes/no).
… showing file sent to MT and returned result.
… showing pre-mt file. Searches for "translate" tag, adds in meta-tag for translator to see translate = no in example shown.
… goes to show source code of translated page, searches for "translate" tag, they are not found, they have been cleaned and removed from the text.
Felix: so after translation, the translate tags are removed?
yes.
Pablo: Showing example with translate global rules. All nodes are translated but their children are not translatable.
… shows example being translated. Nodes are translated, children are not. Showing the tags are being parsed correctly.
Marcais: Are the whole spans (including those with translate = no) through translate system?
Pablo: Yes, all tags are sent through MT
Pedro: The reason is that beind the MT engine these can be the post-editors. The PE may need to seem them for contextual reasons. All data is sent to MT regardless of whether text is to be translated or not.
Marcais: So its just kept as is in the translation (translate =no)
Pablo: Yes. Shows the output that is translated and that which is not.
… flicks between source code and webpage. Shows tags (translate) have been removed.
<Yves_> Question: In the global rule it looks like the XPath expression have no namespace, so how they can work with HTML? Is that related to the issue you mentionned for Namespace in HTML5 earlier today?
… shows domain example. Global rules declare domain. Tag, pointer, mapping.
… multiple "economic" domain mappings are added.
Pedro: These domains are needed by the MT engine.
Pablo: shows the text being translated. Meta-data (domain) has been removed.
… Shows global domain and a different engine based on the domain.
<Jirka> Yves, you are right I was going to ask the same question.
… same rule applies, but different engine is used.
felix: asks a question from Yves and Jirka about domain rule…
Jirka: Selector is not completely right for HTML5 where all elements are placed in HTML namespace.
… I hope that this is not a final version, just a temporary one?
Pablo: No problems, I'll change
it. But will require some work.
... Shows another global rules element. Three different
selector for elements wrt translate yes/no
… different domain.
… When we use domain meta-data we divide the document into different parts. What is sent to the MT engine is shown with tags in place and ITS domain mapping.
… We divide the document into three parts.
… we send three requests to the MT engine.
… it takes longer to translate three as opposed to one,
Pablo: Shows final example applying domain meta-data to different nodes. What is sent to the MT system has multiple domain tags added to it. When page is translated it takes a while…
Pedro: This negativly effects the performance of the MT engine. We must take into account how the ITS usage adds to MT costs.
Milan: How many requests to the MT engine?
Pablo: around 15.
David F: The overhead of multiple fragments depends on the impl of the MT. Some engines may perform better than others.
Pedro: This is how it works, you can balance different engines etc, but this effects the price to the client.
Pablo: as you can see a cache system is present for translation requests.
Milan: Whatabout creating a file per domain and sending three requests, one for each file.
Pablo: Maybe it can be done but its easier to send one file.
Pedro: We are also using PEditors, we have to perform with a CAT standard system.
Pablo: Showing caching system
which results in faster performance.
... Closes demo.
Yves: a general comment about domain value, when we do mapping we dont mention Casing of the domain. Should they be case-sensitive? Should we edit the spec to show this?
Felix: Are keywords case-sensitive in HTML?
jirka: keyword values are case-sensitive.
?: Maybe present in the algorithm, a lower case mapping
<scribe> ACTION: Yves to add a step regarding the lowercasing of the domain data category [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action17]
<trackbot> Created ACTION-271 - Add a step regarding the lowercasing of the domain data category [on Yves Savourel - due 2012-11-08].
david F: our general impl is insensitive to case sensitivity.
… in contract with MT provider you may need to address this but not in our mapping.
Yves: Does it mean the value we return is lowercase as well or should they stay as we get them?
David F: should be lowercase
<fsasaki> I assume "STEP 5: Return the resulting string." would be "STEP 5: lowercase the resulting string and return it."
Marcis: general comment, on translation in general, whether to not send something through an MT system and whether it can be specified to be passed through MT engine or not. Sending a string through MT which is not translated could overload the MT speed. Q: Whether its possible to differentiate to send through an MT system and no translate or do not send through MT system.
David Lewis: Problematic - locale filter allows you to not pass through MT for specific locales. What was being shown was stuff that was not MT'd as it was subsequently PE'd
Marcis: Some things you do not want to translate, or passed through MT engine.
Pedro: Very impl dependent. Depends on what behaviour you want. If something is not translated it matters for analysis of the translation.
Marcis: If you know some text is important for MT and some is not can you handle both?
Pedro: This is a problem with the Lucy MT provision.
… Matrex SMT works with chucks of data, so maybe we can see whether there is an improvement but we dont know yet.
Pedro: Different techniques are used but we need to know how SMT performs.
David F: In ITS we have translate, term, disam which can all be combined to make a business case. But these problems are business issues between LSP and MT provided. We should not spend time on this discussion.
Pedro: This is both true and false. We need to test various contracts between LSP and MT provider and this should be documented.
David F: Agrees but not now.
David L: Has no translate mark-up come up with your clients? Are they interested in this in terms of costs per words?
Pedro: The business model of this contract is not based on that.
<pnietoca> You can find the DEMO link here http://www.w3.org/International/multilingualweb/lt/wiki/Use_cases_-_high_level_summary#More_Information_and_Implementation_Status.2FIssues_7
Felix: Showing screen. 30mins to go, would like to take Bert's presence here to ITS 1.0 uses XPATH 2.0 uses extraction. Shows example of translate data cat, which can be used globally with absolute selectors. 2 subjects, 1: defining selector, 2: using CSS selector at level 3. Think this is the right thing to do based on discussions with Bert. Might be various selectors only relevant to CSS for XPATH.
<pnietoca> and here https://www.w3.org/International/multilingualweb/lt/wiki/Online_MT_Systems_Use_Case_Demonstration#Use_Case_Demonstration
<pnietoca> might be down sometimes depending on the developing
… shows a rule example where node is selected based on evaluated expression with parameter. Having the CSS selector is easier for developers who may not know CSS in detail.
… we dont have anyone who has yet said they would implement CSS selectors. This is feature-risk as no implementor yet avaliable.
… however during break bert said people are working on CSS selectors to XPATH. We could say we support a translation mechanism from CSS selectors to Xpath. CSS level 3.
Jirka: I thought this was already done?
… the code should exist.
Felix: The developer implemented in python, people are working with other frameworks. Wanted to bring this up while bert is here.
… nice to have mechaism to use CSS selector and this to be translated in XPATH. Is this something to specifiy in the Spec, do we point people to libaries etc?
Dave L: If there was mapping from CSS to XPATH do we not need to mention CSS as part of the normative spec?
Felix: If mapping could be part of the spec without being used directly it would be good.
Cocomore?: people are maybe more familiar with CSS selector so would be better to allow users to use CSS selectors. makes more sense if these classes can be selected more easily.
Felix: There is a use-case
… any other thoughts on this?
<mhellwig> s/Cocomore?/kfritsche
Jirka: Is someone working on in-browser implementation.
Yves: I think so as some impl are using CSS. If we go the CSS path what does it change for the test cases? Do we need 2 test-cases for both XPATH and CSS?
Dave L: This question has been asked before.
David F: Doubling the test-suite.
Felix: Just a conversion step, conversion test-case.
… other option is we drop the support for CSS selectors. 1) Full direct support, 2) convergence support or 3) we drop it
Jirka: Who asked for this?
Milan: Maybe it was Phil R?
Dave L: Interoperability is help full in passing file from one system to another.
Felix: Another solution: Dont change anything in the spec, not make it a feature and see if it can be done in an impl
… we would not formally specifiy this step.
… define the functionality without using it formally.
Jirka: keep CSS with reserved word for query language.
Felix: If we dont change anything we put this feature at risk, see if someone comes up with solution next year.
David L: Increases volume of test suite but maybe not the implementation work
Bert: You can only select elements and not attributes..
Jirka: So CSS test cases would be smaller. In CSS the selector is not case sensitive. HTML not case sensitive, XML would be.
…Some algorithm to transform CSS to XPATH would not be easy as case vs. case-insensitive problem.
Felix: For XML there is no strong use-case
… only for thsoe using HTML
Bert: There is no official mapping but everything but the pseudo elements can be mapped to XPATH.
Jirka: It depends on user-interaction
David L: Therefore these things may need to be removed.
Jirka: No these user-interactions will never be applicable
Felix: For the time being no conclusion on this issue but if we dont change anything and no-one impl it may cause an issue. We need to mark this as a feature at risk.
David F: For people to understand the commitment they are making this should be mapped onto the google spreadsheet.
Felix: I will take an action to explain this and see take-up
… this is an opportunity to recruit more rule authors who can use CSS selectors as opposed to learning XPATH.
<fsasaki> ACTION: felix to make sure that css selectors are marked as feature at risk in the draft, and explain the rational in a mail [recorded in http://www.w3.org/2012/11/01-mlw-lt-minutes.html#action18]
<trackbot> Created ACTION-272 - Make sure that css selectors are marked as feature at risk in the draft, and explain the rational in a mail [on Felix Sasaki - due 2012-11-08].
Felix: Topics you want for the agenda tomorrow.
… we have joint meeting with HTML, planning for next year, XML Prague for example. Last workshop in Nov, plans for 2014
… editorial discussions for specification.
… please see doodle poll for a few more people to commit to editing meetings.
<fsasaki> http://doodle.com/heh7k59h7vkvnv88
Jirka: Will the next meeting be in Prague?
Felix: Yes, 23th, 24th of
January.
... Any other topics? Issues?
David F: XLIFF mapping task-force meeting should happen this week. We said we'd circulate this yesterday but we didn't. Can we schedule for tomorrow? Propose 4pm French time.
… 3 pm
Felix: 2-3pm free time tomorrow for implementation discussions.
Jirka: Who will go at 9am to joint meeting with HTML?
Felix: Hope all with join.
… 9am we meet with the HTML group for 30 mins.
Felix: I have asked for a review from the HTML WG on the section on HTML and ITS written by Jirka. Wanted to ask Frederick a demo which Yves gave in Prague. People see what happens to the HTML.
… would be good to have a 5-10 min demo.
Jirka: Closer to 5min
… @felix do you have examples of HTML markup in your slideS?
Felix: Yes I do.
Jirka: Should also show HTML
validator for the WG. Simple, example driven
presentation.
... Our goal is to get a commitment to get a review from
them.
Felix: Looking for a commitment for a contact
Tadej: Plan to discuss tool-info data category tomorrow, would be helpful. I would have some comments on this, if we could discuss before lunch.
Felix: 2 big topics, tool-info
(issue 41) too many global rules (issue 51) others are
smaller.
... So we meet at 8.30 tomorrow here and at 9 we go to the
other room
... Closes the meeting.