I am trying to understand the content-biasing. Somehow it has no effect on our installation.
Stepping through the code learns me that the query-string did change: the modified boost values like
&bq=type:audio^1.0 show up, but have no effect on the search-results' sort-ordering or the $result["score"] value. the score remains exactly the same, no matter how high or low a bias is set.
Could it be cached somewhere on the solr server? (clear cache has no effect, code does not hint at caching it). Is this a known bug? Or an issue with the
Update: Enabling debugQuery=on shows me an explain:
1.7849035 = (MATCH) fieldWeight(body:assen in 1952), product of: 1.0 = tf(termFreq(body:assen)=1) 4.0797796 = idf(docFreq=900, maxDocs=19600) 0.4375 = fieldNorm(field=body, doc=1952) ","eskrkz/node/4622":" 1.1041226 = (MATCH) fieldWeight(body:assen in 1682), product of: 1.7320508 = tf(termFreq(body:assen)=3) 4.0797796 = idf(docFreq=900, maxDocs=19600) 0.15625 = fieldNorm(field=body, doc=1682) ","eskrkz/node/8881":"
The field=body worries me. Why does it only search in body? Why not just any text? And why
An example querystring: