Hi all
I have a question on the implementation of phrase-based and hierarchical
based reordering models in Moses.
According to Galley and Manning (2008)
<http://www.aclweb.org/anthology/D/D08/D08-1089.pdf> , in the phrase-based
model, orientation is set to M if an adjacent phrase pair lies at (s−1,u−1)
in the alignment grid. However, the implementation in Moses is like this:
if((connectedLeftTop && !connectedRightTop) ||
((it = inBottomRight.find(startE - unit)) != inBottomRight.end() &&
it->second.find(startF-unit) != it->second.end()))
return LEFT;
This means if there's an alignment at (s-1,u-1), the orientation will be M.
But existing an alignment at that position does not mean we can extract a
phrase there.
So why implementing like that?
Chees
Liangyou
--
Liangyou Li
PhD Candidate
ADAPT Centre
School of Computing
Dublin City University
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support
