A tool for domain adaptation and data selection for SMT was implemented by Amir Kamran (SLPL Lab. <https://staff.fnwi.uva.nl/k.simaan/research_all.html>, ILLC <https://www.illc.uva.nl/>, University of Amsterdam <http://www.uva.nl/en>) based on
[Hoang Cuong and Khalil Sima’an. <https://www.illc.uva.nl/People/show_person.php?Person_id=Hoang+C.> Latent Domain Translation Models in Mix-of-Domains <http://www.aclweb.org/anthology/C14-1182> on Computational Linguistics]. The developed tool is available at the following Github repository: <https://github.com/amirkamran/InvitationModel> https://github.com/amirkamran/InvitationModel Description The Invitation based data selection approach exploits in-domain data (both monolingual and bilingual) as prior to guide word alignment and phrase pair estimates in a large mix-domain corpus. Accurate estimates are obtained for the probability P(D|e,f) of every mixed-domain sentence pair (e,f) being in-domain (D=1) or out-of-domain (D=0), be used to rank the sentences in mix-domain according to their relevance to in-domain corpus or used directly as model features when extracting phrase pairs. The re-implemenation was conducted at ILLC (Institute for Logic, Language and Computation, University of Amsterdam) https://www.illc.uva.nl in part within the project "Data-Powered Domain-Specific Translation Services On Demand" <http://www.stw.nl/nl/content/data-powered-domain-specific-translation-services-demand-dataptor>, supported by the grant "STW Open Technologieprogramma <http://www.stw.nl/nl/content/open-technologieprogramma>". - Amir Kamran is supported by STW project DatAptor <http://www.stw.nl/nl/content/data-powered-domain-specific-translation-services-demand-dataptor> grant nr. 12271 - Hoang, Cuong is supported by EXPERT ITN project <http://expert-itn.eu/>. - Sima'an, Khalil (2014) is partly supported by NWO Vici project <https://www.illc.uva.nl/AbouttheILLC/Activities/grants.php> grant nr. 277-89-002.
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
