Real clustering is for a web search engine impossible except of you
have unlimited hardware resources.
However as Jerome suggest there is a search result clustering plugin.
If you are more family with Math and algorithms you will find this
article interesting:
http://www.stanford.edu/~taherh/papers/scalable-clustering.pdf
But as mentioned calculating the most important words of a document
and make a new query is the way that is the best solution for now.
Stefan
Am 21.12.2005 um 00:48 schrieb Daqing Zhao:
I think clustering the documents would be a solution and just
recommend
other documents in the same cluster. Is there a clustering
algorithm in
nutch? May be very expensive to calculate.
Daqing Zhao
On 12/20/05, Victor Lee <[EMAIL PROTECTED]> wrote:
Getting the term vector should be easy, but when you said
calculation, is
it a simple comparision of all term vectors, or is it whole
another beast?
Stefan Groschupf <[EMAIL PROTECTED]> wrote: No, nutch has not such a
functionality.
The quick and dirty solution to implement this would extracting the
term vector from the original document, calculate (there would be
different algorithms) somehow the most important terms for this
document and just do a query with these terms.
HTH
Stefan
P.S. Contributions are every-time welcome. :)
Am 20.12.2005 um 04:48 schrieb Victor Lee:
Hi,
Does Nutch's search result show "similar pages" like Google? I
went to Modzex.com which is using Nutch but I don't see "similar
pages" in its search result.
Many thanks.
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
---------------------------------------------------------------
company: http://www.media-style.com
forum: http://www.text-mining.org
blog: http://www.find23.net
