I'm running the example Solr install with a custom schema.xml and
solrconfig.xml. I'm seeing some unexpected results for searches using
the SpellCheckerRequestHandler using the onlyMorePopular option.
Namely, searching for certain terms with onlyMorePopular set to true
returns a suggestion which, when searched for in turn itself, returns
a suggestion back to the original term.
For example, a query such as:
http://localhost:8983/solr/select/?
q=Eft&qt=spellchecker&onlyMorePopular=true
returns:
<response>
−
<lst name="responseHeader">
<int name="status">0</int>
<int name="QTime">2</int>
</lst>
−
<arr name="suggestions">
<str>oft</str>
</arr>
</response>
And a query for
http://localhost:8983/solr/select/?
q=oft&qt=spellchecker&onlyMorePopular=true
<response>
−
<lst name="responseHeader">
<int name="status">0</int>
<int name="QTime">2</int>
</lst>
−
<arr name="suggestions">
<str>Eft</str>
</arr>
</response>
It seems that "onlyMorePopular" should be an asymmetric relation. I
thought perhaps it might actually be implemented as a >= instead of a
strict >, making it antisymmetric and perhaps explaining this result
as a popularity tie. However, taking a clean copy of the example
install, adding entries into the spellchecker.xml file, then
inserting and rebuilding the index results in onlyMorePopular cross-
recommendations as above even when I've created a clear popularity
inequality between similar terms (e.g. adding two docs with
word="blackkerry" makes it more popular than the existing
"blackberry" doc, but each is suggested for the other).
I checked the defaults list for the spellchecker requestHandler in my
solrconfig.xml, and it didn't specify a value for onlyMorePopular. I
added a default value of true and restarted Solr, but that has no
effect. I've also tried using Luke to inspect the spell index, but
I'm not sure exactly what to look for. I'd be more than happy to
provide any details which might assist others in lending their
expertise. Any insights would be very much appreciated.
Thanks,
Justin Knoll
