Lists Home |
Date Index |
From: Matt Gushee [mailto:email@example.com]
>All I know is that it's a .NET component that does handwriting
>recognition. You don't go to Microsoft presentations expecting to find
>out *exactly* what a technology is, do you?
Thanks. Good point.
>As for the training, you may be right. But I believe the presenter
>claimed there was no need for it in most cases.
That would surprise me if it works well. Handwriting recognition
is an awfully difficult problem without some drastic simplifying
assumptions and those tend to lead to noise. Consider how badly
a lot of OCRing works and that for printed glyphs.
Costs accrue downstream from input sources if the input has to
be validated by different parties. We all know the deal having
dealt with forms-based systems, datatype validation, rule validation
and so on. If the entry device *only* speeds up the physical entry
of data and doesn't vette, the fabled cost savings are lost and
costs can even go up. This is typical problem of field reporting
systems: the officer doesn't want to spend more time reentering
data the system says is bad, and the organization doesn't want to
bear the costs of downstream data scrubbers. Turning a $1000 laptop
PC into a pad and pencil doesn't seem like progress unless that
pencil is awfully clever.