52nlp's Learning Notes » POS

Moses Support Digest:POS LM

52nlp — Wed, 30 Dec 2009 12:52:10 +0000

[Moses-support] POS LM

There is pos.lm of the target language in factored model training. I want to know the steps involved in preparing the POS.lm and the kind of input parameters altogether.

-Doren

Re:[Moses-support] POS LM

Hi Doren,

please read the tutorial on factored models:

http://www.statmt.org/moses/?n=Moses.FactoredTutorial

-phi

Re:[Moses-support] POS LM

Hi Doren,

I’ve used SRILM to generate POS LMs. The LM, as you might expect, needs to be training on a corpus consisting of sequences of POSes instead of sequences of surface forms, e.g. instead of

The cat sat on the mat

the corpus should contain

DET N V P DET N

or whatever.

Furthermore, the set of POSes is probably small as vocabularies go, so smoothing methods that rely on counts-of-counts, such as Kneser-Ney, are inappropriate. The SRILM website’s FAQrecommends Witten-Bell discounting (command line option ‘-wbdiscount’) for such cases. (See question C3, answer (b) at the FAQ.)

Also because the vocabulary is small, you can get away with using
higher-order n-grams than you would use for a surface LM.

Other than that, it’s the same as preparing a surface LM.

Regards,
Ben

NOTICE:This is digested from the Moses-support mailing list, which supports for the moses SMT decoder.

Moses Support Digest:POS translation

52nlp — Mon, 07 Dec 2009 12:46:11 +0000

[Moses-support] POS translation

Dear All,
I am looking for using moses to translate some sentence that contains POS. for example : I want a fly for the first of november I want to keep my POS by using the “-xml-input pass-through” option like :

input : I want a fly for the first of november
output : je veux un vol pour le premier novembre

but I didn’t find any clue about the probabilities that can be associated to my POS in the moses translation process.
Is there any one can give me some information about it ?

Thanks,

Christophe

Re:[Moses-support] POS translation

Hi,

if I understand this correctly, you want to preserve the XML tags verbatim and only translate the content. The easiest solution would be to strip out the XML, translate the text, and then using the phrase-alignment (and word alignment within phrases) to determine the positions where the tags should be inserted in the output.

This is the method that is used in our web page translation demo:

http://demo.statmt.org/web.cgi

-phi

NOTICE:This is digested from the Moses-support mailing list, which supports for the moses SMT decoder.

Moses Support Digest:Building POS language model with SRILM

52nlp — Wed, 25 Nov 2009 12:52:10 +0000

[Moses-support] Building POS language model with SRILM
Hi,
The Moses manual recommends using the following switches when building a language model with SRILM:
　　-interpolate -kndiscount
I assume this recommendation applies specifically to surface-string language models. For a part-of-speech language model, KN-discounting is inappropriate because it is based on counts-of-counts, and the counts-of-counts for POSes are odd in that there are very few POSes that occur only once or twice in a given corpus.
Are there particular switches that are recommended for building a POS
language model with SRILM?
Regards,
Ben Gottesman

Re: [Moses-support] Building POS language model with SRILM

Hi,
you are correct that for POS LMs the lower order n-gram counts are very different and smoothing is less relevant.
You could train a 7-gram LM with Good Turing smoothing for the lower order n-grams and Kneser-Ney for the higher order n-grams.
I have done this occasionally.

-phi

NOTICE:This is digested from the Moses-support mailing list, which supports for the moses SMT decoder.