On Mon, Aug 23, 2010 at 4:01 PM, Dwayne Bailey <dwa...@translate.org.za> wrote: > On Fri, 2010-08-20 at 17:52 -0700, The Language Techie wrote: >> Dwayne: >> >> In Twitter you mentioned that Pootle core code is in place for real-time TM. >> Can you give a little more details, and when this functionality coming? > > We've just got 2.1 out the door and will probably focus on bug fixes for > the next little while. After that we'll be looking at our roadmap. > >> This is a wish list description of what I consider to be real-time TM >> support: >> >> (1) Project initiators can upload a TMX TM file to the project to apply the >> TM matching to all the files for that project. > > As apart from the files being translated? You mean a separate TM right. > >> (2) At the "translate" interface, each source segment will have a list of >> fuzzy-match suggestions based on the active TM for the project. The score >> of the fuzzy match (1%-99%) or exact match (100%) will be displayed. The >> difference in the source text and the text being considered for translation >> will be highlighted for the translator to see the difference right a way. > > We do this on Virtaal, have a look and tell us what you think. This > would probably be the approach that we'll take. > >> (3) The translator can choose to see fuzzy matches at a pre-configured >> level, such as 0% (no fuzzy), 50%, 60%, 75%, etc. This should be a user >> preference setting. > > While setting a threshold is nice, a really low threshold has no real > value and loads the server quite considerably.
IMHO below 60% is not useful at all. I even suggest to show only the first five top results at most. For short strings (less than 10 words) I only suggest to show 85% or upper matches, and in larger strings (paragraphs and long phrases) you can show results with a matching of 60% or more. These are my suggestions after using Lokalize TM matching feature during several months, I still am searching the way to setup the TM matching feature on Virtaal. >> (4) After one translation is committed or suggested (by pressing the >> button), that translation goes into the active TM pool immediately, so that >> it can become a 100% match or a fuzzy match for the next sentence, without >> waiting until the entire project is completed or when the TM is updated. > > With 2.1 all our strings are in the database so that should become > easier to do. But the issue of load comes up again. Matching is pretty > expensive. > >> For example, >> >> EN "The quick brown fox jumps over the lazy dog." => zh-TW "敏捷的棕狐狸跳過懶狗。" >> >> As soon as this translation is made, and "committed" or "suggested", if at a >> later time, or in the next sentence, the translator sees: >> >> EN "The quick brown foxes jump over the lazy dog." >> >> They will see in the fuzzy-match section that >> >> EN "The quick brown fox jumps over the lazy dog." => zh-TW "敏捷的棕狐狸跳過懶狗。" >> >> is a high fuzzy match, with a score of about 75-80%. If this TM match is >> only a suggestion, it will be nice to show that. Any TM match should show >> information like creation-user, change-user, creation-time, change-time, >> etc., if available. > > We probably won't do the TM approach that loads a 100% match into your > text. We prefer users to make their own selections. > >> This real-time fuzzy match behavior is what commercial CAT tools offer >> already. > > Sure, so do we in Virtaal. Its pretty easy to do on a desktop tool. > But matching a lot of strings across lot of languages and projects will > be where we need to see how well we can scale. > >> (5) At any time during the project, the active TM can be downloaded through >> TMX format. It should contain all committed and suggested translation. Or, >> the inclusion of suggested translations in the download should be an >> option. The download TMX should contain creation-user, change-user, >> creation-time, change-time, etc. meta data. >> >> (6) When the user chooses to download files for offline translation with >> Virtaal, the relevant part of the active TM (over the chosen % threshold) >> should be downloaded along with the files, so that in Virtaal, users will >> see exactly the same fuzzy matches as they would see in Pootle online. The >> only difference is that, translations made by users offline in Virtaal will >> not be uploaded to Pootle server in real-time to benefit other translators. >> >> Hope my description of the TM functionality makes sense to the team, and I >> look forward to hearing about this feature! > > Yes, it does indeed and is very much inline with our own thinking. > > -- > Dwayne Bailey > Associate Research Director +27 12 460 1095 (w) > Translate.org.za ANLoc +27 83 443 7114 (c) > > Recent blog posts: > * Localizing Mac OS X strings files using open source PO editors > http://www.translate.org.za/blogs/dwayne/en/content/localizing-mac-os-x-strings-files-using-open-source-po-editors > * What's new in Virtaal 0.6.1 > * Localisation: How we guess the target translation language in Virtaal > > Firefox web browser in Afrikaans - http://af.www.mozilla.com/af/ > African Network for Localisation (ANLoc) - http://africanlocalisation.net/ > > > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by > > Make an app they can't live without > Enter the BlackBerry Developer Challenge > http://p.sf.net/sfu/RIM-dev2dev > _______________________________________________ > Translate-pootle mailing list > Translate-pootle@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/translate-pootle > ------------------------------------------------------------------------------ Sell apps to millions through the Intel(R) Atom(Tm) Developer Program Be part of this innovative community and reach millions of netbook users worldwide. Take advantage of special opportunities to increase revenue and speed time-to-market. Join now, and jumpstart your future. http://p.sf.net/sfu/intel-atom-d2d _______________________________________________ Translate-pootle mailing list Translate-pootle@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/translate-pootle
