Stemming and stop word filters can corrupt search results in many cases. The two analyzers appear linked to the language setting. Language value "none" disables text analysis completely. It would be nice to have "plain" with just a generic tokenizer.