2

I'd like to ask questions about personalized search. I'm about to design/implement a personalized search with Lucene. I did some googling about that, but didn't seem to find module/tools ready to use with Lucene. So I guess I need to implement one on my own. I would like to know if there's any tools available so that I don't reinvent the wheel.

If nothing is available, it's fine too. But I would like to know if there's any suggestion on how to implement it. My current idea is about adding an extra bit-vector or bloom filter to each document. Each bit represents the relevancy of the document to a user or a group of similar users. The relevancy info is derived from data mining algorithms. I am not exactly sure if this is the optimal solution (cheap to implement at the same time).

Any suggestion/comment will highly appreciated! Thank you!

gre_gor
  • 6,669
  • 9
  • 47
  • 52
ycl
  • 86
  • 1
  • 8

1 Answers1

1

You could add a field representing the matched groups / users, and boost the individual values. For example, the field group for the document "horses and butterflies" would have the boosted tokens group:"schoolgirls"^10 group:"bikers"^-1 while the document "road horses" would have the boosted tokens group:"schoolgirls"^2 group:"bikers"^10. Searching for horses would return road horses first in case you are a biker, and butterflies first in case you are a schoolgirl.

wesen
  • 631
  • 6
  • 10