Analyze differences in relevance ranking, zope.textindex vs. pgtextindex

Estimated Effort: 2d

Goals
========

- Figure out why search results are different
- See in which cases one is better than the other
- Compile a list of suggested improvements
- Later tasks will analyze performance/scalability impacts

Details
=========

- Get both running side-by-side using OSI data
- Find the N most recent text searches for OSI
- Compare results, trying to gauge which would have been "better"
- Do analysis to see why each are yielding different results
- See how the text extraction, synonyms, stopwords, stemming, weighting, and
  other parts of the equation compare
- Propose changes in the index plugin protocol to improve relevance
- Write down any places (e.g. "Don't index the profile photo") that might yield
  better quality search results
- Include in the list any oddball ideas for boosting relevance (dynamic weighting
  based on count of tags, whatever)
- See if results are different/better when faced with foreign languages/characters

Question
==========

1) Are we doing an implicit prefix search on the advanced search page? I think
there was once a theory that it should return the same results yielded by the
LiveSearch. Which is, in hindsight, a dumb idea.

2) Does the pgtextindex URL use system ispell dictionaries?