Prior art

[mjcollin]

https://www.google.com/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=looking+for+duplicates+in+gbif&start=10

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3677402/
http://update.specifysoftware.org/6400/relnotes.html (SGR)
http://wiki.filteredpush.org/wiki/2013Oct09
http://dev.gbif.org/issues/browse/DM-231

[Bouteloua]

Just talked to the Filtered Push people. They're clustering collator names and collator numbers, and then letting the user clean it up after that point. They know if one of the fields is NaN this will create issues, but this is a quick an dirty "fix". Bob Morris is currently working on the data mining aspect...