Teleform 3.0 is powerful software for automatically converting customized paper forms to computer data
Stan Miastkowski
The fabled ``paperless office'' has become one of the running jokes of the PC revolution. PCs were supposed to eliminate paper; but, if anything, they've perpetuated it. According to market research firm BIS Strategic Decisions, most Fortune 500 firms use more than 10,000 different forms. Add the many and varied forms from the tens of thousands of small businesses, and the numbers quickly become mind-boggling.
Most of that information on paper must eventually be transformed into data that computers can use. Currently, that's largely done by legions of data-entry people, most of whom spend endless dreary days on this mind-numbing exercise. What's needed is a way to automati
cally take forms, read the information, and transform it into computer-readable data. Cardiff Teleform for Windows is a unique package designed specifically to do that--to directly transform hand-entered data on forms into standard database formats.
Teleform is in version 3 and has added some features, including toolbars and customizable grids, that make it easier to use than its predecessors. More to the point, its character-recognition abilities have been honed to make them impressively fast and accurate. What sets Teleform apart from stand-alone OCR programs are its forms-centric design and its direct export into database formats.
Teleform uses a constrained design for forms, which increases the odds of error-free recognition. As with all OCR programs, however, Teleform's success rate is still far from perfect--many forms require ``cleaning up'' before they can be entered in the database, as we found in our tests. The range of potential uses for Teleform is wide indeed, although the software
seems to have found its biggest proponents in organizations that gather data by fax (or mail) from large groups (for a typical example, see the text box ``The BYTE Hardware Poll'').
Serious Hardware for Serious Software
Teleform is a large and complex application that requires serious hardware resources. This is especially true when it's running at full tilt, because Teleform multitasks several programs concurrently. Installation is easy, but, practically speaking, you'll need at least a 486 with 8 MB of RAM and about 40 MB of hard disk space to get started (more if you'll be storing many forms and responses). For high-volume forms processing, a multiuser version is available that does workload balancing by spreading the forms-processing chores among multiple networked PCs (I used the single-user version for this review).
Processing faxed-in forms is the heart of Teleform's functionality, and instead of reinventing the wheel, Cardiff has included SoftNet's popular FaxWorks program. FaxWorks l
inks directly to the Teleform modules and also gives you an excellent fax program for other faxing chores. FaxWorks supports most fax modems, and the installation utility automatically senses, identifies, and tests your fax modem.
Teleform also includes extensive scanner support, and I had no problems interfacing it to a Hewlett-Packard IIcx flatbed unit. Cardiff offers optional drivers for high-speed scanners with automatic paper feeding. You will, however, need to make sure your PC hardware is compatible. I wasn't able to get a high-end Fujitsu scanner to work, because the Teleform driver required an Adaptec SCSI adapter and wouldn't work with the Future Domain SCSI board in my test system.
Form Design
Teleform's Form Designer is well designed and easy to use. If you've used a desktop publishing program, most of Form Designer's components will be familiar. If you haven't, creating a form won't require extensive expertise. You just need to plan ahead to design efficient and useful forms.
All Teleform forms include large ``cornerstones'' in each corner. The program uses these, along with an automatically created Form ID block, to identify received images (faxed or scanned) as Teleform-created and to tie them to form parameters that you create during the design stage.
Form Designer gives you almost unlimited control over type fonts, sizes, and attributes. One thing it lacks, however, is an easy way to import graphics, such as company logos. You can't just import a PCX file, for example, although you can cut and paste from any Windows program that creates BMP, TIFF, and WMF files. Instead, you must open your graphic in a separate application (e.g., Windows PaintBrush), copy it to the Clipboard, press Alt-Tab to open Form Designer, and paste it in.
Object Objectives
Teleform offers a variety of objects, each designed for capturing different types of data (see the figure ``The `Content' of a Form''):
-- Optical Mark Recognition (OMR) -- ``Color the dots'' entry.
-
- Optical Character Recognition (OCR) -- Recognition of preprinted text (e.g., a form ID number or serial number).
-- Constrained Handprint Recognition -- Handprinted entry of alphanumeric characters; only one character per box is allowed.
-- Unconstrained Handprint Recognition -- Handprinted entry without the one-character-per-box constraint, usually compared against a user-defined dictionary or database.
-- Complex BasicScript Validations -- Teleform contains its own BASIC-like language that lets you perform complex operations on received data.
-- Image Capture--TIFF images (e.g., signatures).
Choosing objects for your particular application is crucial to form design. Which objects are best for which types of data is well covered in the Teleform manual. The more constraints you place on the data, the more accurate its recognition will be.
Once you choose an object, all you need to do is drop it on the form. Then you fill out a dialog box that gives the data a field I
D (for database entry) and check boxes for other attributes. Although it looks complex at first, Teleform makes the process easy, especially with a preview box that shows you what an object looks like before you paste it into a form. Teleform also has a library of predefined objects (such as name and address fields) that you can drop into any form.
Dictionaries and Databases
It quickly becomes apparent that most fields in a form have a limited universe of possible entries. Thus, one way to increase recognition accuracy is to provide Teleform with a list of possible values. There are two ways to do this: dictionaries and databases. A dictionary is simply a list of possible words. You can enter your own or import any text file. Database validations are more complex to set up, but they operate on fields instead of individual words.
Once you have created your form, you have only a couple more steps before you can distribute it. First you must choose the database format that the forms will be expo
rted to after they are recognized. Teleform supports a wide range of formats, including dBase, Paradox, Access, Excel, and ODBC (Open Database Connectivity), the ``middleware'' that connects to large client/server database systems. Finally, you have to ``activate'' the form, notifying Teleform that the form is finished and ready to be received.
The Shock of Recognition
Once you've distributed the forms (e.g., via fax or mail), you're ready for the reception/recognition phase, which is largely automatic. With FaxWorks up and running, it receives forms and feeds them to the Reader, the real brains of Teleform (you can also scan forms directly into the Reader). The Reader identifies, recognizes, and passes each form to the Verifier, which flags forms that need additional correction. You can manually correct them or choose to have errors ignored and the data fed directly to the database.
This approach, used with forms that are highly constrained to increase accuracy, is often necessary in high-vo
lume operations. The Verifier's configuration menu also offers extensive options for fine-tuning, such as the Confidence Threshold, which tells Teleform how much effort to make in creating substitutions for characters it's not sure about.
True to Form
Ultimately, most of Teleform's shortcomings are the same as in all recognition software--no one has yet figured out how to reliably spot the patterns in handwriting and unconstrained handprinting. But Teleform lets you minimize its weaknesses and exploit its strengths by using constrained objects as much as possible. As a rule, forms with no handprinting objects will be recognized close to 100 percent of the time, and they can handle most people's information-gathering needs. If you must use handprinted entry, the lower recognition rate isn't likely to wipe out the still- substantial savings in time and labor.
At first glance, Teleform's $1495 price tag may seem high, but the program is an eminently useful, cutting-edge application that pushes t
he edge of technology. Despite its underlying complexity, Teleform is rock solid and surprisingly easy to set up and use. Careful planning during form design is crucial to achieving the highest accuracy. But if you need to gather data from paper forms and convert that data for computer analysis, the bottom line is that Teleform can pay for itself in no time flat.
The Facts
Teleform for Windows 3.0 $1495
(upgrade from version 2.0, $295; multiuser version,
two- and five-user, $1495 per user)
Cardiff Software, Inc.
531 Stevens Ave., Bldg. B
Solana Beach, CA 92075
(800) 659-8755, (619) 259-6430
fax: (619) 259-6450
Illustration: Unconstrained text entry (top) and careless handprinting that ignores example text (bottom) can create problems for Teleform (note the best guess in the New Value box at the upper left corner of each screen). Recognition accuracy can be improved by creating a custom dictionary or database lookup.
Figure: The
`Content' of a Form
The tested form contained three types of recognition objects and a predefined ``Example'' object. Handprint is generally recognized less reliably than optical marks, though recognition improves when printed characters are constrained to boxes. Recognition also improves when users follow the model in the ``Example'' object.
Stan Miastkowski is a BYTE consulting editor who has been writing about networking and communications technology for over 16 years. He is co-author of Windows for Workgroups Bible (Addison-Wesley, 1993). You can contact him on the Internet or BIX at
stanm@bix.com
.
Collecting Facts from Fax
