You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-user@lucene.apache.org by Brian Whitman <br...@variogr.am> on 2007/02/02 01:13:43 UTC
searching by field's TF vector (not MoreLikeThis)
I'm looking for a way to search by a field's internal TF vector
representation.
MoreLikeThis does not seem to be what I want-- it constructs a text
query based on the top scoring TF-IDF terms. I want to query by TF
vector directly, bypassing the tokens.
Lucene understandably has knowledge of the cosine dist of these
vectors -- does it expose it in such a way that I can query the topN
results from a field's TF vector?
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org
Re: searching by field's TF vector (not MoreLikeThis)
Posted by Brian Whitman <br...@variogr.am>.
On Feb 1, 2007, at 7:13 PM, Brian Whitman wrote:
> I'm looking for a way to search by a field's internal TF vector
> representation.
>
> MoreLikeThis does not seem to be what I want-- it constructs a text
> query based on the top scoring TF-IDF terms. I want to query by TF
> vector directly, bypassing the tokens.
After looking around the archives & Lucene's code & etc, my
assumption is:
Lucene does not use the entire TF vector space in any search. There
is no tree search or other log-n search mechanism built into Lucene.
TF cos dist is using for scoring, once the search space is reduced
from the occurrence of terms in query from the inverted index, and
then it's a foreach document operation.
If this is incorrect, please let me know.
-Brian
---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org