Solr is currently deployed on several servers on our image sharing site. We have 10 million images, 1/4 million is added every month.
So far, Solr has done a very good job of selecting search results, but we believe that there is an opportunity to improve sorting / ranking. We believe that including click metrics in the ranking will provide significantly improved results.
We are currently collecting click data through MongoDB. We register how many times a click on a click, and how many times it is displayed for each term. For example:
[image id], [search query], [click-through rate]
"00000001", "banana peel", "0.1565"
"00000001", "banana", "0.0216"
"00000001", "monkey banana", "0.0087"
What we want to do is find a way to include this data with clicks on search words in our Solr rankings. The more the image was clicked for the same term, the higher it will be ranked. We still managed to find a way to do this cleanly.
We would like to download data from the outside and not have it as part of the Solr index - since we want the data with a click to be almost in real time and would like our Solr directory to be huge too.
Any ideas or thoughts will be very much appreciated!
source share