Given the current development of the Lucene suggest module, I believe it
makes sense to refactor the Solr Suggester out of the existing SpellCheck
Component into its own.

https://issues.apache.org/jira/browse/SOLR-5378 [patch uploaded]
It would be really great if you could look at the current
input/output/config of the Component and provide some feedback! (the
details are in the jira description). The patch has a working
implementation of the new Component (SuggestComponent and SolrSuggester)

The idea is to give the Solr Suggester the flexibility in terms of
response, processing (distributed & normal) and pluggability for utilizing
the new features of the Lucene counterpart. The new Component will allow:
  - input (dictionary) pluggability [users can choose from the Dictionary
implementations available, including utilizing the new expressions module
in lucene for term weights]
  - Allow us to add new ways to build the suggesters [move computation from
query time to index]
  - Have desired input and output formats suitable for the suggester.

I plan to do the following:
  - add tests & documentations
  - add support for all the lucene suggesters [FreeTextSuggester is the
only one without support]
  - add additional lookup Factories to allow users to filter out
suggestions based on categories instead of doing query time processing

Thanks in advance,

Areek Zillur

Reply via email to