From: Per T. <per...@op...> - 2012-09-05 14:17:46
|
Hi, how do I proceed to work efficiently with improving da-se? Where to start? What steps have I to go through? Yours, Per Tunedal On Wed, Aug 15, 2012, at 10:40, Jacob Nordfalk wrote: 2012/8/14 Per Tunedal <[1]per...@op...> > Likewise, people who work on MT are recommended to begin with > translating into their mother tongue :) (that's also the reason for the > state of da→sv, no Swedes have worked on it). I see. There is no end to the work to be done by a Swede! True! What exactly has to be done before releasing the pair da - sv? I have already told you, but in a way that not apparent. Sorry for that. Here I'll try to explain what happens and where the problems are. Dansk til svensk virker ikke -> (Analysis of all possible interpretations of the words using Danish monidict. Works OK.) ^Dansk/Dansk<adj><pst><un><sg><ind>/Dansk<n><nt><sg><ind><nom>$ ^til/til<adv>/til<pr>/til<cnjadv>$ ^svensk/svensk<adj><pst><un><sg><ind>/svensk<n><nt><sg><ind><nom>$ ^virker/virke<vblex><pres><actv>$ ^ikke/ikke<adv>$ -> (tagging, ie choosing which one of the possible interpretations of a word is the correct one. Probably broken as the tagger hasnt been trained on a Danish corpus) ^Dansk<n><nt><sg><ind><nom>$ ^til<pr>$ ^svensk<n><nt><sg><ind><nom>$ ^virke<vblex><pres><actv>$ ^ikke<adv>$ -> transfer and use of bilingual da-sv dictionary. Broken as 1) 'ikke' maps to both icke and inte and 2) and 'virke' is not found in the bilingual dictionary. ^Danska<n><ut><sg><ind><nom>$ ^till<pr>$ ^svenska<n><ut><sg><ind><nom>$ ^@virke<vblex><pres><actv>$ ^icke<adv>/inte<adv>$ -> generation of Swedish text. Probably OK. Danska till svenska \@virke #icke You should really take a look with Apertium-viewer. As da - sv unfortunately has been disabled as its not production-ready, you have to download the pair from SVN and compile it first. Mikel, I think Per is a perfect example of why unreleased pairs should also be available for download: Using the easy tools is the best way for a potential contributor to be able to judge the existing work. > > What matters is whether it's different in nb and sv … Well, as I've said before; I plan to reuse as much as possible from the pair sv - da. It would be perfect if you could work a little on improving da->sv to get easy experience, and then shift to your main goal. I and Francis were the mentors for sv->da and we'd be happy to answer all questions. If you pass Copenhagen then we could also meet in person to get you up and running. Jacob -- [2]Jacob Nordfalk [3]javabog.dk Androidudvikler og -underviser på [4]IHK og [5]Lund&Bendsen ----------------------------------------------------------------------- ------- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. [6]http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Apertium-stuff mailing list [7]Ape...@li... [8]https://lists.sourceforge.net/lists/listinfo/apertium-stuff References 1. mailto:per...@op... 2. http://profiles.google.com/jacob.nordfalk 3. http://javabog.dk/ 4. http://cv.ihk.dk/diplomuddannelser/itd/vf/MAU 5. https://www.lundogbendsen.dk/undervisning/beskrivelse/LB1809/ 6. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ 7. mailto:Ape...@li... 8. https://lists.sourceforge.net/lists/listinfo/apertium-stuff |