I just had the same problem. It turns out I called train-model.perl with only the -hierarchical option, which is not enough. My problem was solved by adding -source-syntax and -target-syntax.
2010/6/24 Lucia Specia <[email protected]>: > Hi all, > > I managed to extract rules from a parsed parallel corpus (thanks to > Hieu!), but some of them contain XML strings that I believe should not > be there, for example: > > ( [pu] ||| <tree [pu] ||| ||| 0.0526316 1 0.025 1 2.718 ||| 19 40 > , [pu] ||| </tree> [pu] ||| ||| 0.357143 1 0.0028169 1 2.718 ||| 14 1775 > , [pu] ||| <tree [pron-pers] ||| ||| 0.166667 1 0.00056338 1 2.718 ||| 6 1775 > </tree> <tree label="np"> <tree [np] ||| a base de cimento [np] ||| > ||| 1 1 1 1 2.718 ||| 0.5 0.5 > > I've checked the XML and it seems to be alright. > > Is this something expected? > > Thanks a lot, > > Lucia > _______________________________________________ > 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
