Stimulation Initiative for European Neural Applications
Esprit Project 9811
Automatic Handwriting Recognition
Company background
Helvetia Krankenkasse is the largest Swiss health care insurer with
1.4 million members. It is a division of the Swiss Care organization.
The entire organization has a total of 2.4 million members (30 % of
the Swiss population). It processes six million claims annually, 25,000
each business day. In Switzerland there are 1,300 regional offices
and 11 administrative centres.
The problem
For years 18 data processing clerks processed the incoming claims on-line
at a central location. Because one data processing clerk could handle only a
maximum of 150 claims per hour, Helvetia had to contract out some 3 million
claims per year to an external data-entry agency, at a cost of $ 0.75 per
claim. The benefits of a faster and more efficient
claim processing procedure are clear.
Neural network application
Document Access developed for Helvetia a system for automatic
handwriting recognition.
The software of system is based on neural network technology which
enables a high level of recognition percentages and, in contrast
to hardware based solutions, provided optimum flexibility. The system
subsequently underwent intensive training in a large assortment of Swiss
handwriting styles, so it isn't important who completed the forms; virtually
all of the numbers and letters are recognized. Only about three out of 10,000
characters are recognized incorrectly.
The system has been installed in January 1994.
In practice the system works as follows. Helvetia employees
place the claim forms in the scanner.
The forms are scanned simultaneously on both sides and then automatically
recognized. The 168 data fields on the form -- 75 % numerical, 10 % alpha-numerical
and 15 % alpha-characters -- are processed automatically by the software.
Whenever necessary, the typists enter only a few manual corrections; during this process
a number of recognized characters are verified on the basis of information
which appears on the screen.
Benefits
The processing capacity of a clerk increased from 150 to 400 claims per hour.
Helvetia no longer has to turn to other data-entry agencies for claims
processing. The total savings for Helvetia are $ 2,300,000 per year. The
investment costs have been recouped in
eight months.
Generalization
The key feature of this application is the use of a neural computing
to recognize human handwriting. This is a task which conventional computing
has found virtually impossible to tackle. Wherever large numbers of
handwritten data - for instance from order forms, tax forms or cheques - are
manually keyed into a computer, this technology has
obvious advantages.
Contact persons
H. Altman, Helvetia Health Insurance, Ringstrasse 12, 8600 Dübendorf,
Switzerland. Tel +41 1 824 5220, Fax +41 1 820 2830.
A. A. Veenhof, Document Access BV, Westersingel 101, 3015 LD Rotterdam,
The Netherlands.
Tel +31 10 436 66 64,
Fax +31 10 436 68 44.
References for further reading
A.C.R. Hogervorst et al. Character Recognition Using Neural Networks.
In:
Neural Networks: Artificial Intelligence and Industrial Applications.
Proceedings of the 3rd Annual SNN Symposium on Neural Networks, Nijmegen,
The Netherlands, 14-15 September 1995, Nijmegen. B. Kappen and S. Gielen (eds).
Springer-Verlag, London.