On 3/24/07, Paul Novitski <paul@xxxxxxxxxxxxxxxxxxx> wrote:
At 3/23/2007 07:27 PM, Richard Lynch wrote: > > In this case, the OP has an existing list of names he wants to > > de-capitalize, not an ongoing stream of new names from people who > > might be trained. > >The solution remains: > >Hire a human. > >The computer will never get accurate enough. > >The exception might be if you are dealing with MILLIONS of names, >where a filter would pay off. You'd still need human review and a >validation process that involved human oversight to a significant >percentage. > >I do not think the OP has millions of names. Richard, If you'd take the time to read this thread, you'll see that many of us have pointed out the impossibility of correcting name capitalization 100%. You'll also see that the OP has indicated that even an imperfect solution with capitalization errors would be better in his situation than the current all-caps condition of the list, and that therefore a software solution would be helpful. There is usually more than one solution to a given problem. In this case, as in many others, there appears to be an inverse relation between cost and error rate. Depending on goals and budgets, one may choose a higher error rate and a lesser cost over a smaller error rate at a higher cost. In this particular case, it's my opinion that even human operators would not be able to reduce the error rate to zero unless one could afford to pay them to contact some of the people individually to find out how they spell their names. There simply are no rules that apply across the board, whether applied by machine or flesh. Anything but personal interviews is just informed guesswork. Regards, Paul
Sure, even a human can't know if the name should be MacDonald or Macdonald, McDonald or Mcdonald. Unless of course he's contacting the one who entered this name. Tijnema
__________________________ Paul Novitski Juniper Webcraft Ltd. http://juniperwebcraft.com -- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
-- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
