On Feb 23, 2008, at 3:05 AM, Otis Gospodnetic wrote:
Aha, good example, Sean. What's the explanation? Note that doing:
http://www.google.com/search?q=abdur+choudhury
offers this alternative:
http://www.google.com/searchq=abdur+chowdhury
And that the number of hits is approximately the same in both cases
and that Google is smart enough to search for and highlight
chowdhury even when the search was for choudhury.
Google's spelling corrections/suggestions are driven off of massive
query (refinement) logs. Solr's suggestions are based on the index
field content.
They probably are based off both the logs and index, right? One thing
that might be of interest is Karl Wettin's adaptive spell checker
patch in Lucene, which I think attempts to learn as it goes, plus uses
the index, but I haven't looked at it at a low level.
-Grant
