If you could move to 3.x and your linked item boosts could be
calculated offline in batch periodically you could use an external
file field to store the doc boost.
a few If's though
On 8 June 2011 03:23, Jeff Boul zbou...@hotmail.com wrote:
Hi,
I am trying to figure out options for the following problem. I am on
Solr 1.4.1 (Lucene 2.9.1).
I need to perform a boost on a query related to the value of a multiple
value field.
Lets say the result return the following documents:
id name linked_items
3 doc3 (item1, item33, item55)
8 doc8 (item2, item55, item8)
0 doc0 (item7)
1 doc1 (item1)
I want the result to be boosted regarding the foollowing ordered list of
linked_items values:
item2 item55 item1 ...
So doc8 will received the higher boost because his 'linked_items' contains
'item2'
then doc3 will received a lower boost because his 'linked_items' contains
'item55'
then doc1 will received a much lower boost because his 'linked_items'
contains 'item1'
and maybe doc0 will received some boost if 'item7' is somewhere in the list.
The tricky part is that the ordered list is obtained by querying on an other
index. So the result of the query on the other index will give me a result
and I will use the values of one field of those documents to construct the
ordered list.
It would be even better if the boost not use only the order but also the
score of the result of the query on the other index.
I'm not very used to Solr and Lucene but from what I read, I think that the
solution turns around a customization of the Query object.
So the questions are:
1) Am I right with the Query's cutomization assumption? (if so... can
someone could give me advices or point me an example of something related)
2) Is something already exist that i could use to do that?
3) Is that a good approach to use separate index?
Thanks for the help
Jeff
--
View this message in context:
http://lucene.472066.n3.nabble.com/Boosting-result-on-query-tp3037649p3037649.html
Sent from the Solr - User mailing list archive at Nabble.com.
