Hi Prashant
This uses the suffix array phrase table, described here:
http://www.statmt.org/moses/?n=Moses.AdvancedFeatures#ntoc31
I think you can find more about it here:
homepages.inf.ed.ac.uk/miles/phd-projects/*levenberg*.pdf
For hierarchical model, you can use Adam Lopez suffix array
implementation which also works with Moses
http://www.statmt.org/moses/?n=Moses.AdvancedFeatures#ntoc39
On 08/08/2012 10:33, Prashant Mathur wrote:
Hi,
I am working with moses server. There is an option to use training
corpus and alignment file instead of normal phrase table.
Instead of
1 0 0 5 <path to phrase-table>
we can have
8 0 0 3 <train.src> <train.tgt> <alignment>
Am I right? But it seems the way of scoring the phrases is different here.
How does the decoding works in this case? What are the three scores?
Can someone point me to any tutorial on this because I didn't find any
on statmt.org/moses
Thanks,
Prashant
_______________________________________________
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
