Hey folks, I'm giving a talk at ApacheCon in two weeks about how domain
knowledge and/or knowledge of your user base can be used to boost/bias the
scores of documents in Solr search results. Simple examples being things
like: using function queries to boost by numeric fields like date or
popularity; or customizing the tf and lengthNorm functions in your
Similarity based on what "good" documents look like.
http://www.apachecon.eu/schedule/presentation/16/
I would love to be able to highlight some examples beyond things i've
personally worked on, so I wanted to reach out the the entire solr-user
list to see if anyone had interesting examples they'd like to share about
ways they have tweaked the scoring of documents scores based on knowledge
of thier users/documents.
Most of the "classes" of customizations i can think of involve...
* Customized Similarity
* Boost queries or functions
* ExternalFiledField
* QueryElevationComponent
...but don't be shy about sharing a technique you've used that didn't
involve any of those things.
You can feel free to respond to this message publicly with as much/little
details you're willing to share about the application, users,
customizations you used. Or if you'd like to share an interesting use
case anonymously (ie: w/o divulging your name or company) you can send it
to me privately.
Thanks.
-Hoss
