Hi Hieu, Thank you very much for the details and for making your thesis available.
What I was looking for was that when I use train-model.perl with --source-syntax and --target-syntax parameters, which of these methods are executed behind the scene. According to your answer, it looks like something taken from all those work, and not solely based on one of them (if I'm not mistaken. Thanks again, Rasul. Quoting Hieu Hoang <[email protected]>: > hi rasul > > there is an implementation of the GHKM extract algorithm in > training/phrase-extract/extract-ghkm > the program > training/phrase-extract/extract-rules > can extract generic string-to-tree and tree-to-string grammar. It can > also extract SAMT-style grammar. > > Since these algorithms are reimplementations of existing research no-one > written about them. You should just read the original papers on > SAMT/GHKM/syntax models etc. > > In all these cases, the decoder algorithm is the same (CKY+). Phil > Williams has also implemented the Scope 3 algorithm by Hopkins et al. > > The Moses syntax decoder was also written to do another tree-to-string > model which I wrote about in > http://aclweb.org/anthology-new/W/W10/W10-1761.pdf > A better explanation for it is in my thesis > www.statmt.org/~s0565741/ddd.pdf > > On 30/03/2012 19:47, [email protected] wrote: >> Hi Barry, >> >> Thanks for reply. I have actually read that paper. According to it, if >> I understood correctly, the string-to-tree rule extraction is based on >> DeNeefe et al. (2007), which is in turn based on GHKM (Galley et al., >> 2004). However, for me it's not clear about tree-to-string and >> tree-to-tree models; whether they are using the same principle or >> based on other techniques? >> >> Regards, >> Rasul. >> >> Quoting Barry Haddow<[email protected]>: >> >>> Hi Rasul >>> >>> This paper is probably a good start, and should give you pointers >>> to further >>> literature. >>> >>> http://mastarpj.nict.go.jp/IWSLT2009/proceedings/TP_4_hoang.pdf >>> >>> cheers - Barry >>> >>> On Friday 30 March 2012 12:13:[email protected] wrote: >>>> Hi all, >>>> >>>> I was trying to find out what methods (publish work) the different >>>> syntactic modeling approaches (tree-to-string, string-to-tree and >>>> tree-to-tree ) in Moses are based on. I was wondering if somebody >>>> could help me with references of such published work. >>>> >>>> Regards, >>>> Rasul. >>>> _______________________________________________ >>>> Moses-support mailing list >>>> [email protected] >>>> http://mailman.mit.edu/mailman/listinfo/moses-support >>>> >>> -- >>> Barry Haddow >>> University of Edinburgh >>> +44 (0) 131 651 3173 >>> >>> -- >>> The University of Edinburgh is a charitable body, registered in >>> Scotland, with registration number SC005336. >>> >>> >> >> _______________________________________________ >> Moses-support mailing list >> [email protected] >> http://mailman.mit.edu/mailman/listinfo/moses-support >> > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support > _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
