Re: [Palmkit-users-jp] Sorry for bothering you again

[Akinori I. <ai...@fw...>]

Hello,

Khan Sakeb wrote:

>I think,at first I'm gonna create a content-word-only
>corpus and a content-word LM P_C(c|a,b).Then,for
>predicting the probability P(w_f|a,b),I should use the
>entire corpus(containing both Content words and Function
>words). Now, for predicting the following probability,
>which corpus should I use? 
>P_C(w_c|a',b')(1-Σ_{w in F} P(w|a,b))
>  
>
If you have models that calculate P_C and P, you don't need any
additional corpus to calculate the probability above.
(If you mean a test set, it should be a corpus independent from
the training corpus, and the test set should include both content
words and function words.)

>And,again I'm gonna ask a very funny question.Is there any
>way that I can predict this probabilities using Palmkit
>only(I mean by making a few changes in input or output
>options)??Or,do I have to take any kinda programming
>approach.Not to mention,I dont have that much confidence
>in programming(PERL) also.So,it will be a great relief for
>me if you kindly give me any hint of solving this problem
>using Palmkit(if possible).
>  
>
Palmkit provides with functionality to train an ordinary trigram. You
have to
write some program to choose content words from the entire corpus
(Palmkit does not have such command). Ask your supervisor about the
programming issues.

>Finally,let me apologize once again for bothering you.My
>heartiest gratitude is there always for you.Thank you very
>much.
>
>With Regards.
>
> 
>
>--------------------------------------
>Start Yahoo! Auction now! Check out the cool campaign
>http://pr.mail.yahoo.co.jp/auction/
>
>  
>