Predefined Topic lists

Manar

Hello everyone..
I have a question.
1- When I have predefined topic lists, which contains some words to extract the suitable topic of each Arabic documents.

Cosine similarity is considered a good solution for this problem?

or latent Dirichlet allocation (LDA) ?

Please, could you guide me to do that in rapidminer? 
Thanks.

Telcontar120

This is an interesting question.
@mschmitz is the resident expert on LDA (well at least he has written the operator) but I am pretty sure that is not going to help you here because I don't think you can feed the LDA algorithm a predefined set of topics.

So I am not actually sure what the best way to accomplish this would be. I guess you could put together a wordlist with the words for each predefined cluster and then try to build a polynominal classification model but that might not give you the output you really want.  @mschmitz do you have another approach you would recommend here?  

P.S.  I don't think the language is really an issue, it has more to do with the structure of the problem.