On Fri, Mar 30, 2012 at 2:08 PM, Thomas Schoenemann <
thomas_schoenem...@yahoo.de> wrote:
> Hi everyone,
>
> there is quite a discussion about hierarchical and syntax-based
> approaches here at the moment. For lack of time I could not follow it
> closely, so I hope this question contributes to the discussion rather than
> being superfluous: in terms of {tree/string} -> {tree/string}, how would
> you classify Hiero? My understanding of tree -> {string/tree} is that you
> have a source tree given, which is not the case for Hiero. And
> string->string does in my understanding not involve CFG-like rules at all.
> So Hiero would be string->tree. But this solution is entirely based on
> eliminating the alternatives. I'd prefer to have an expert come up with a
> constructive explanation.
>
> Thanks much!
> Thomas (currently University of Düsseldorf)
>
> P.S: MERT now has an optional positivity constraint.
>
Thomas,
You are correct. Hiero is string->tree.
The input is a plain string (as opposed to a parsed tree).
The output is a special type of tree, where the only types of nonterminals
are S and X. Reading off the leaves of the output tree in order gives you
the translation.
SAMT uses the same CKY+ decoding algorithm that Hiero does. The difference
is that SAMT and similar grammars use a full set of linguistically
motivated nonterminals in their SCFGs. The result is the same, though -
input is a string, output is a tree.
Cheers,
Lane
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
