lcrocker@nupedia.com wrote:
Would it be possible, or desirable, to have a system which does a search for one particular error at a time, eg "recieve", and corrects all the pages it finds in batches, of say 20, to avoid loading the server?
I think it should be a general policy that we should not allow any automatic process to alter content. The value of Wikipedia is that real human beings with understanding and judgment have edited the pages.
What if there were a page about common English mispellings that had "recieve" there intentionally? Or perhaps it could be someone's name, or a foreign word used in context. I remember having a Vietnames co-worker named "Teh" who was constantly running into the problem of spell checkers changing his name to "The".
I see your point. I was thinking of something that only tackles one particular mis-spelling at a time, and not a full spell-checker for that reason; but I see that even a batch replace of all "recieve" could break things like: 'a common misspelling of "receive" is "recieve"' 'Recieve is a city in Brazil" (a feasible typo for Recife) and so on.
Something that gives a list of, say 10 occurences, with context, and asks for check-box confirmation of each one might be okay -- but I think it should search for only one particular mis-spelt word. Overall, the task of checking spelling and grammar has to be a human one.