Hi, Arjun Upadhyaya wrote: > Hieu Hoang <hieuho...@...> writes: > > >> using other search algorithms like GA or swarm is an interesting idea. >> >> you may want to look at this paper which uses random sampling for decoding >> http://homepages.inf.ed.ac.uk/s0343799/conll09_arun_et_al.pdf >> Its operators would be similar to mutation and crossover operators you >> would need for GA. >> > Hey Hieu Hoang, > Thank you very much for re directing me to that paper which > uses random sampling for decoding sentences on to the target language. So I > just want to ask you whether I should use MOSES as an evaluating toolkit like > BLEU or can I incorporate GA's within MOSES itself. > Also are there any other papers which deal with the same > subject??? > we did something similar, but we used a greedy algorithm to improve existing translations:
http://www.mt-archive.info/TMI-2007-Langlais.pdf which is similar to what Germann et al did a while ago with their ReWrite decoder: http://www.isi.edu/natural-language/projects/rewrite/decoder.pdf Depending on how you use GA, I guess these would be fair baselines to compare against. Best regards, Alexandre Patry _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
