Hi,
I wonder whether the word alignment information in the phrase table is used
in decoding? Such as ||| (0) (1,2,3) () ||| (2) (1,2) |||. I think moses
does not use the lexicon, so how does these information be used?
Thanks.
Qin
___
Moses-support maili
nope. the alignment information is not used by the decoder in the main
trunk.
_
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Qin Gao
Sent: 05 August 2008 14:08
To: moses-support@mit.edu
Subject: [Moses-support] Alignment information in Phrase table
Hi,
I wonder wheth
Oops, forgot to CC the list.
> From: "John D. Burger" <[EMAIL PROTECTED]>
> Date: August 4, 2008 13:30:30 EDT
> To: [EMAIL PROTECTED]
> Subject: Re: [Moses-support] decoding: reordering only
>
> Sanne Korzec wrote:
>
>> Is there a way to force the moses or pharaoh decoder, to use a
>> certain se
Hi -
I'm still trying to debug my differences between old and new versions
of Moses, which (for us) use SRILM and IRSTLM respectively. My
current puzzle is over the very different sizes of the language
models resulting from SRILM and IRSTLM - the latter has 5 times as
many 5-grams, for in
by default the srilm prunes singletons
Miles
2008/8/5 John D. Burger <[EMAIL PROTECTED]>
> Hi -
>
> I'm still trying to debug my differences between old and new versions
> of Moses, which (for us) use SRILM and IRSTLM respectively. My
> current puzzle is over the very different sizes of the lan
Miles Osborne wrote:
> by default the srilm prunes singletons
OK, that's good to know. But when I prune the IRST LM, I still get
lots =more= 4-grams than the SRI LM, but lots =fewer= 5-grams
(although less than a factor of two in either case).
But perhaps I'm a bit in the weeds here ... :)
you want to also check that ngrams are not getting pruned by probability (in
addition to counts)
this whole business is a bit on the murky side and the only reason i know
about it was when i was writing a disk-based version of ngram-count a year
or so back
Miles
2008/8/5 John D. Burger <[EMAIL P
Miles Osborne wrote:
> you want to also check that ngrams are not getting pruned by
> probability (in addition to counts)
Yes, in fact, this:
http://www.speech.sri.com/projects/srilm/manpages/ngram-count.1.html
seems to suggest that pruning is done based on not changing
perplexity very m
it has been a while since i looked at this, but look at this (good-turning):
*not pruning*
[rydell]miles: ./ngram-count -lm /tmp/test2.lm -order 3 -gt1min 0 -gt2min 0
-gt3min 0 -text ../../../mt/diskbased-l m-training/temp.txt
warning: discount coeff 1 is out of range: 5.55654e-17
warning: disc
Hi all,
IRSTLM uses simpler smoothing methods than SRILM.
In particular, improved kneser ney smoothing with
SRILM uses corrected frequencies for lower order
n-grams, while IRSTLM does not. (This indeed results
in less n-grams.) The reason is that
introducing corrected frequencies makes it hard
to
2008/8/5 John D. Burger <[EMAIL PROTECTED]>:
> I'm starting to think it's a lost cause to try to get one LM
> implementation to act very much like the other. Thanks for the
> insights, though!
I also spent some time unsuccessfully trying to exactly match the
SRILM toolkit's output. Aside from the
actually there are two parts here --building large LMs and deploying
them.i currently have a summer MSc project looking at using Hadoop and
Hbase to do this Google-style. this really does use a cluster of machines,
for both parts. in either case, building them on-disk with a single machine
or
Hi,
I installed Ubuntu Linux 8 with successful execution of
training-factored-model. I am also tried to install Ubuntu Linux 7.10 on
coLinux, and the result is successful also.
I don't have any particular idea why in solaris the training is fail. One
different thing I learn is the MERT version wh
13 matches
Mail list logo