Hi, the tree-to-tree model is a pretty natural extension of the string-to-tree and tree-to-string model.
You need to provide a parse tree of the input, and grammar rules have to match nodes in the tree. At the same time a tree structure in the output is built during translation. This is not a very popular method for syntax-based machine translation at this point, because rule application is very constraint. Hence, there is little literature on it. The implementation in Moses is straight-forward: it performs non-terminal label checks as it would do for both string-to-tree and tree-to-string. -phi On Tue, May 15, 2012 at 3:34 AM, <[email protected]> wrote: > Hi all, > > I have asked the similar question before for syntax-based approaches > in Moses in general. While I now have some idea of which publihsed > work tree-to-string and string-to-tree implementation is based on, I > would like to know about tree-to-tree model. As far as I know there > only a few tree-to-tree work in literature, and there is no clue which > one is the basis for Moses implementaion of the model. > > Thanks and regards, > Rasul. > _______________________________________________ > 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
