[Iglu-java-users] Re: Help

[Travis B. <tr...@us...>]

Adriana,

I'm sorry I didn't reply to your last email.  Choosing
an indexing vocabulary is a good question and it
depends on your data set.  IGLU is design not to make
those decisions for you.  After you have trained the
vector creator by passing the documents through it,
you can make two method calls to control the vocabular
and size of each index.  setLimitTopN(int) will tell
the vector creator to only use the top N most
frequently occurring terms in the indices generated. 
setMaxSize(int n) will tell it to only return indices
with the top N most highly ranked terms for each term
vector created. 

In my experience, I find that setLimitTopN is not
terribly effective because by limiting the vocabulary
size in that way, you run the risk of dropping out
terms that help differentiate documents.  setMaxSize
is usually good to use because you don't want your
vectors to be too large.  I'd try using a
setMaxSize(25) and see if you get good performance
with that.  

Does this answer your question?  

Travis

--- Adriana - IG <adr...@ig...> wrote:
> Dear Travis Bauer,
> 
> Do you know how can I get the indexing terms from
> the docs.? How many 
> terms does need it have.?
> 
> I used all classes from the IGLU to index the
> document. And I have good 
> results. I only don´t know how many terms select to
> Index list.
> 
> 
> Thank you.
> 
> Adriana.
> 
>