Fwd: ACTION-154 Look into ISSUE-26 post editing data categories (IGNORE PREVIOUS POSt)

** Apologies - Please ignore previous post which was sent unfinished - 
here's the completed version.

Hi All,
One existing input to this topic is Pedro's post;
http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Jun/0050.html

This identified some parameters related to post-editing:

  * UTS Ratings (Utility, Time and Sentiment)
  * Utility (relative importance of the functionality of the translated
  * content).
  * Delivery Time (speed with which the translation is required).
  * Sentiment (importance on brand image).
  * Expiration level.

For me these all fell into the class of localization job parameters 
similar to ISO TS 11669. We had agreed should not fall under ITS2.0 but 
should be  left for now to be driven by Linport and ISO, and perhaps 
revisited with them once we get into best practice mode next year, as 
summed up by arle;
http://lists.w3.org/Archives/Public/public-multilingualweb-lt/2012Jul/0030.html

I dug around for other activities on post-editing. One possible 
requirement is MT confidence scores, which could be important meta-data 
to convey from an MT service to a posteditor using a CAT tool. But that 
probably makes it more of a target for XLIFF than ITS - David, Yves, I'm 
not sure if that is on the agenda there?

One specialization of this is that might impact content markup in a way 
relevant to ITS is sub-segment confidence scores, which are readily 
extracted from SMT engines like MOSES (I've see presentations from 
AsiaOnline that displayed this for posteditors). Again, though that 
still might be more appropriate for XLIFF, especially the inline markup 
group?

As a general point, there are three new FP7 projects i know of in the 
broad area of PEMT, addressing issues such as subsegment postediting or 
how to convey confidence scores to users/crowdsource translators, as 
well as work in CNGL. See;
ACCEPT: http://www.accept.unige.ch/index.html
CASMECAT: http://www.casmacat.eu/
MATECAT: http://www.matecat.com/matecat/the-project/

This indicates to me this is still an area of active research and 
therefore not stable enough to consider as ITS requirements:

Declan may well have some further views here.

On balance therefore i don't see any compelling requirements for fresh 
postediting data categories specifically coming from the WG. We could 
possibly add an optional 'confidence-score' attribute to the local 
translationAgent data category, but I'm not sure we'd have a real use 
case for this yet since we don't know how to really interpret such a 
score, apart from as a local confidence differentiator.

So i recommend taking no action for now and closing ISSUE-26.

cheers,
Dave

p.s. thanks to Johann Roturier of Symantec for kindly sharing his 
knowledge in this area