Finding singular versions of a word in Sunspot/Solr - ruby-on-rails

Question

I have a Rails+Sunspot application and I'm working on configuring it so that searching returns the singluar version of the query. For instance:
I want a search for "cookies" to return something named "cookie". Currently my Sunspot search returns "cookies" but not "cookie" (singluar).
I've made some customizations to Solr's schema.xml, adding solr.EdgeNGramFilterFactory to provide more flexibility but EdgeNGramFilterFactory doesn't suite this case as it only allows matches when the query is a substring of the result's name. My understanding is EdgeNGramFilterFactory will return "cookie" when the user searches for "co", "coo", "cook" or "cooki", but not a superstring of "cookie" (ie: cookies). Simply put, this is because "cookies" is not a substring within "cookie".
I've tried adding all three of Solr's build-in stemming factories but to no avail. You can see one commented out in my schema.
In schema.xml, the relevant field looks as follows:
<fieldType name="text" class="solr.TextField" omitNorms="false">
<analyzer type="index">
<tokenizer class="solr.StandardTokenizerFactory"/>
<filter class="solr.StandardFilterFactory"/>
<filter class="solr.LowerCaseFilterFactory"/>
<filter class="solr.EdgeNGramFilterFactory" minGramSize="2" maxGramSize="15" side="front"/>
<!-- <filter class="solr.EnglishMinimalStemFilterFactory"/> -->
</analyzer>
<analyzer type="query">
<tokenizer class="solr.StandardTokenizerFactory"/>
<filter class="solr.StandardFilterFactory"/>
<filter class="solr.LowerCaseFilterFactory"/>
</analyzer>
</fieldType>
I supposed I could singluarize the user's query but I would rather not touch their query before it hits Solr.
You can play with this here: http://staging.zisboombah.com/parent/food_guide/?search=cookie. Try changing the query between "cookie" and "cookies".
Any tips on how to do this in Solr would be greatly appreciated!

Answer 1

The solr xml options are ordered. You want the stemmer to come before the ngram filter, so that you ngram-ize cooki, rather than stemming c, co, etc.
Combining filters in this way may lead to some odd results, mostly depending on how aggressive your stemmer is. You should definitely add the stemmer to the query analyzer, but that will mess with your autocomplete.
A better solution: use a copyField to make independent text_stemmed and text_autocomplete fields. Then search using an OR query over both fields.

Answer 2

Like Kyle mentions, you probably want to use more text field types for each of these different use cases.
Here's an example of mine:
schema.xml
<schema>
<types>
<fieldType name="text" class="solr.TextField" omitNorms="false">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
<filter class="solr.StandardFilterFactory"/>
<filter class="solr.LowerCaseFilterFactory"/>
</analyzer>
</fieldType>
<fieldType name="text_en" class="solr.TextField" omitNorms="false">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
<filter class="solr.StandardFilterFactory"/>
<filter class="solr.LowerCaseFilterFactory"/>
<filter class="solr.PorterStemFilterFactory"/>
</analyzer>
</fieldType>
<fieldType name="text_stopwords" class="solr.TextField" omitNorms="false">
<analyzer>
<tokenizer class="solr.StandardTokenizerFactory"/>
<filter class="solr.StandardFilterFactory"/>
<filter class="solr.LowerCaseFilterFactory"/>
<filter class="solr.StopFilterFactory" words="stopwords.txt" ignoreCase="true"/>
</analyzer>
</fieldType>
<!-- ... -->
</types>
<fields>
<!-- ... -->
</fields>
<copyField source="*_text" dest="text"/>
<copyField source="*_texts" dest="text"/>
<copyField source="*_textsv" dest="text"/>
<copyField source="*_textv" dest="text"/>
</schema>
Sunspot modeling
Using the copyField directive can save some setup work in the model. However, Sunspot uses those text declarations to decide which fields to keywords-search by default, so I like to include distinct text invocations that use :as to specify the full Solr document field name.
searchable do
text :name, stored: true, default_boost: 10
text :name, as: 'name_text_en'
text :description, stored: true
end