I am currently having an issue where entering technical search term that is not English, but rather is "English-like" does not return the proper results. Our site is built to support a development community that uses my company's software products, so many of the terms searched for are not real words, but are completely valid terms in our proprietary lingo. The behavior I am seeing is that the search results returned seem to be the result of a correction taking place on the search term to make it actual words.
For example, a configuration option a developer may set in our systems is called useSSL. There are specific nodes that should be returned for this term, based on the content of those nodes.
Instead, I do not get back the pages containing this term, but rather get back pages with the term "uses SSL" or "use SSL" (note the white space in between these terms, and the expansion of "use" to "uses").
I have seen similar behaviors where the search terms entered are a camel cased mashup of two real words, such as messagesLog, which returns hits on message log, messages log, but NOT on messagesLog.
I have disabled the "Enable spellchecker and suggestions" option, but still see this behavior. My questions then, are:
1) Once the spell checker is disabled, do I need to re-index to prevent the above-described behavior?
2) If this is not the proper way to suppress this behavior, what is?
Thanks in advance.
Comments
Comment #1
pwolanin commentedYou will have to spend some time with Solr itself understanding the indexing and searching process - this is not something controlled at the level of the Drupal module, but rather in the Solr configuration files (of which we provide a reasonable starting set, but it's not suitable for all cases).
Likely you need to use protwords or disable stemming. Also consider using common grams indexing and look at the way token are generated on case change.
see: http://wiki.apache.org/solr/AnalyzersTokenizersTokenFilters
and related pages there.
Comment #2
maestrogeek commentedThanks for the info! We'll research the above items and link. Appreciate your help.