It is easy to convert a STSG to a weakly equivalent SCFG, though. That's how many people deal with STSG in MT systems, and depending on what you want to do, it may be sufficient.
On Sun, Mar 27, 2011 at 11:38 AM, Hieu Hoang <[email protected]> wrote: > If you're creating a german-to-english system and you only get German in > the output, it's not normal. It should only output german words when > that word is unknown in the translation model. Double check your > translation model and how it was trained. Also, start with a hierachical > model to make sure your setup is working. > > Moses doesn't support STSG and i'm not sure if any open source toolkit > does. It's definitely something that should be attempted > > On 27/03/2011 13:47, Kaveh Taghipour wrote: >> After training a tree-based German to English translation model, I >> used MERT to optimize the weights. A file named 'run1.out' is created >> during tuning and contains sentences in German. Is it normal? As I >> know it has to be in English to be useful in tuning. Is there >> something wrong with my tuning phase? >> >> Also, does Moses decoder support synchronous tree-substitution >> grammars (STSG)? If not, do you know any freely available decoder for >> that? >> >> Best regards, >> Kaveh Taghipour >> > _______________________________________________ > 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
