Hi all, I have trained and tuned a tree-to-tree syntax-based model using Moses successfully, then I also used "scripts/analysis/extract-target-trees.py" to convert the output of decoding into a target tree. The tree format looks like this:
<tree label="Q"> <tree label="Q"> <tree label="Q"> <tree label="Q"> <tree label="Q"> <tree label="Q"> <s> </tree> <tree label="Named"> O0 X0 tom </tree> </tree> <tree label="X"> trusted|UNK|UNK|UNK </tree> </tree> <tree label="Pred"> O0 X0 female n.02 </tree> </tree> <tree label="DRS"> B0 X B0 X0 day n.01 </tree> </tree> </s> </tree> You can see that almost all the tree labels are annotated with “Q” or “X”s. Since I am trying to use Moses to complete a parsing task, what I really want is the parsed tree rather than plain text. So I am wondering if it is possible to get rid of the tree labels like “Q” or “X” and use the syntax annotations (real non-terminals) instead? Thank you for your help! Best regards, Shihao
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
