Getting By Until Real Recognition Is Ready

Editor’s Note: Gene Townsend, who is a vice president at Distributed Micro Systems, Inc., can be contacted at 433 Airport Blvd., Suite 434, Burlingame, CA 94010. DMS has been developing pen-based apps for over two years ago. Initial projects at DMS were based on the GridPad, although development has been extended to GO’s PenPoint, CIC’s PenDOS, and Microsoft PenWindows. This article touches on some application design alternatives that may suffice until true handwriting recognition becomes a reality.

Extensive overselling of “handwriting” recognition continues to plague the nascent pen-based computing industry. Even optimists acknowledge that true handwriting recognition is at least a year or two away. “Handprinting” recognition, however, is here today, although current software for recognizing printing is still too slow for entering large amounts of data and different users will have different levels of success in “printing correctly.”

In almost all cases, good system design can eliminate the need for handprinting data entry fields. Two alternatives to having the computer recognize printed letters by the user exist — look-up lists and graphic fields.

Look-Up Lists

Look-up lists provide choices to the user for selection by the pen. Usually the pen is positioned on the field and a pop-up window appears with a list of possible responses. This technique is a method to provide structured responses when there is a need to update mainframe databases. If a choice of “other” is required to a specific data field, a free-form data entry field can be used. Hopefully the system designer has provided the user with enough choices that the “other” response is not usually required.

The advantages of look-up lists are in the ease and speed of error free data entry. Look-up lists are also easier to understand for the programming staff accustomed to traditional programming platforms. This advantage does not come without some cost. Most important is that it is the system designer, not the user, who is making most of the choices. This is fine in an environment that by its very nature is intended to be confined — an opinion poll that will produce statistics on a limited set of questions, for instance.

Graphic Fields

If part of the power of the pen is in writing responses, and character recognition is not yet good enough for mainstream horizontal applications, how does the system designer provide the necessary tools for the unstructured application? The answer today lies in the use of graphic fields.

One of the best early demonstrations of this approach was a police accident report form application developed by SoftAd (Sausalito, CA) for Grid as a demonstration product. First, the type of intersection was chosen, followed by the placement of graphic objects such as stop signs and lights. All of this was done by choosing the graphic element on the screen and dragging it to the appropriate location. The officer’s notes could then be added as freehand text anywhere on the electronic form. This text is not recognized or translated by the computer, but remains as a graphic object on the screen (an annotation, if you will). The officer, later looking at his notes, can clearly determine just what was written.

Once the system designer chooses the graphic approach, the next question is where does the information go and how is it stored. PenPoint has an excellent store-and-forward engine built into its operating system. Documents can be automatically faxed whenever connected to a phone line. The recipient of the fax might be another PenPoint-equipped computer, an automated file cabinet (computer system with optical disk), or a standard fax machine for human inspection.

Graphic images are not likely to be supported well on existing mainframe systems. This is the primary reason many early systems made limited use of the power of graphics and images. Imaging systems offer the ability to store the graphic information with the character recognition fields limited to the data required for upload to the corporate mainframe and indexes for subsequent retrieval. This goes a long way in eliminating the file cabinets and paper required for many applications today.