Hi,
Since the topic of huge reordering tables appeared I just wanted to let 
you now that I am working at the moment on a much smaller reordering 
table and phrase table representation. The reordering-table is already 
working and could be used within Moses without much problems.
For instance, I have a reduction from 114G to 5.2G for the reordering 
table. Right now however, it needs to be kept fully in RAM. Some 
memory-mapping capabilities are planned. It also needs quite a lot 
memory during construction time, but I will do something with that, too.

If someone wants to take a look or share some comments, it's in the WIPO 
branch, LexicalReorderingDictionary.* (class 
LexicalReoderingDictionaryHashed).

Best,
Marcin

W dniu 03.02.2012 05:50, Wang Pidong pisze:
> Thanks a lot!
> I am going to increase my disk space and try again.
>
> Best wishes!
> Pidong
>
> On 3 February 2012 12:11, Hieu Hoang <[email protected]
> <mailto:[email protected]>> wrote:
>
>     yes, that's typical.
>
>     as a comparison, my 3.9GB (text, gz) reordering table takes up 40GB
>     when binarized.
>
>     On Fri, Feb 3, 2012 at 10:49 AM, Wang Pidong <[email protected]
>     <mailto:[email protected]>> wrote:
>
>         Hi all,
>
>         I am now trying to binarize a big lexical reordering table,
>         *reordering-table.wbe-msd-bidirectional-fe.gz*, whose size is
>         2.4GB after gzip.
>         After about 6-hour running of the *processLexicalTable* tool,
>         the incomplete binarized table (about 68GB) used up all my disk
>         space.
>
>         Does anyone have the experience of binarizing big lexical
>         reordering table?
>         Is my case typical? do I need to increase my disk space then try
>         again?
>
>         Thank you in advance!
>
>         Best wishes!
>         Pidong
>
>
>
>
>         _______________________________________________
>         Moses-support mailing list
>         [email protected] <mailto:[email protected]>
>         http://mailman.mit.edu/mailman/listinfo/moses-support
>
>
>
>
>
> --
> Wang Pidong
>
> Department of Computer Science
> School of Computing
> National University of Singapore
>
>
>
> _______________________________________________
> 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

Reply via email to