On 15/07/2015 20:18, Vincent Nguyen wrote:
>> Hello,
>>
>> Sorry if this is basic but I am trying to get started and something is
>> unclear to me in the manual :
>>
>> Section 2.1.3 recommends (Hieu) Mgiza for Word Alignment and says
>> Moses includes KenLM
>> which makes me understand that IRSTLM and SRILM are optional.
>>
>> Build is fine.
>> Section 2.1.8 Run Moses for the first time is fine
>>
>> Then comes the baseline section 2.3 :
>>
>> Can I skip installing Giza++ since I have installed Mgiza as mentioned
>> before ?
that's fine. I only ever use mgiza too
>>
>> Corpus training is fine.
>>
>> Then if I skip installation of IRSTLM ---since it is optional --- do I
>> need to skip completely 2.3.4 ?
>> is there any Language Model Training to perform if we stick to KenLM ?
you need to create a language model. You can do it with kenlm using the
program lmplz. The program is described here:
https://kheafield.com/code/kenlm/estimation/
We haven't updated the tutorial with more info about lmplz yet, but we
should do at some point
>>
>> Subsequently, if 2.3.4 is not performed, there is no
>> news-commentary-v8.fr-en.blm.en file which means
>> no ability to Train the Translation System.
>>
>>
>> Does the tuning 2.3.6 remain relevant without the above steps ? but in
>> this case what the "working" directory ?
yes, you MUST tune to get good translation quality.
>>
>> I hope my questions are clear but basically how do we follow the
>> Baseline guide without using IRSTLM ?
that's not a problem. It's probably common now use use kenlm in place of
SRILM and IRSTLM for training and decoding.
>>
>> Thanks
>> Vincent
>>
> _______________________________________________
> Moses-support mailing list
> Moses-support@mit.edu
> http://mailman.mit.edu/mailman/listinfo/moses-support
>
--
Hieu Hoang
Researcher
New York University, Abu Dhabi
http://www.hoang.co.uk/hieu
_______________________________________________
Moses-support mailing list
Moses-support@mit.edu
http://mailman.mit.edu/mailman/listinfo/moses-support
