Sure it would be automatic, using algorithms such as "soundex" or "metaphone", or also using the "levenshtein distance".
Actually I did something similar in a previous project, so I feel confident ;-)
It would be interested to have a report of the keywords, grouped by "sound".
For example,
* php my visit
* php my visite
* php my visits
* php my vis
* php my visities
* php visites
All these words would be grouped into a bigger bucket, for example "phpmyvisites".
It would be easy then to see the main 'words' of interest, independently from the spelling or other small changes.
I like the feature, although I wonder how feasible it is.
Would this be automatic?
misteurti