Imagine reading an entire printed page and instantly converting those inert characters into streams of computer-ready data. In 1951, engineer and inventor David Shepard achieved exactly this feat, developing the world‘s first optical character recognition (OCR) device. Dubbed "GISMO", its photoelectric scanner and pattern-matching algorithms foretold today‘s age of automated text digitization. This article charts David Shepard‘s remarkable technical achievement and GISMO‘s outsized influence on modern computing.
An Insatiable Appetite for Building
From childhood, David Shepard reveled in constructing contraptions. He excelled in science and mathematics but was no bookish introvert. Friends described his gregarious, outgoing enthusiasm and appetite for conversation – especially regarding new ideas and inventions. Shepard‘s inventiveness blossomed during WW2, developing cipher machines and encryption systems for Army communications. But he considered these "solutions looking for problems", with his real passion being creation for civil sector applications.
Teaching a Machine to Read
In 1950, Shepard rented attic space in suburban Virginia to construct what he called "an apparatus for reading". Having enlisted longtime friend Harvey Cook Jr. in the endeavor, Shepard set out to build an automatic system for recognizing printed letters and converting text into electronic information. It represented a moonshot effort – no existing technology could yet reliably scan and identify patterns of standard alphabet characters. But over 18 months of painstaking development, Shepard‘s audacious goal moved dramatically toward reality.
GISMO – A Technical Marvel
On December 6th, 1950, the first successful end-to-end demonstration of Shepard‘s reading machine confirmed his breakthrough. Christened the "Geometric Interpreter for Symbolic Manipulation of Objects", GISMO could scan typed text from a page, recognize 23 upper-case letters, and output streams of computer-processable digital code. By 1951, all 26 letters of the English alphabet were supported – a feat once considered virtually impossible by contemporary scientists.
So how did this technical marvel work? GISMO‘s genius resided in Shepard‘s ingenious photoelectric scanning apparatus…
GISMO‘s 1933 vacuum tube amplifier (left) driving the photosensor scanning array (right)
The scanner housed 21 tiny photocells in a horizontal array spanning 5 inches wide – approximately the breadth of a typed line of text. A variable brightness light source would illuminate each letter of printed text, as the page passed across this sensor array.
Depending on the opacity shape of the letter, more or less light would cast onto the photocell detectors. This generated an analog signal representation of the printed letter, amplified by vacuum tube circuitry. GISMO‘s internals housed the real magic however…
Advanced Pattern Recognition – in 1951!
GISMO‘s analog signals fed into clever electronics which cleverly matched signal patterns against ‘templates‘ stored in memory. For the letter "A", it compared captured waveforms against a prototypical "A"waveform shape, distinguishing it from a "B" or other character. This optical pattern recognition converted the signal into a digital letter identity output – astonishingly advanced for pre-transistor 1950‘s technology!
Shepard benchmarked GISMO‘s letter recognition accuracy at over 96% – comparable to modern systems. Processing at 5 letters per second, it reliably ingested typed text into pure machine-readable code. As pages flowed through GISMO‘s hunger scanner each day in Shepard‘s lab, a new era of digitizing the printed word had begun. But could such an unprecedented capabilityfind commercial backing?
Validating a Breakthrough Invention
Shepard knew GISMO represented no mere attic tinkering project. Partnering with scientist William Lawless Jr., the two incorporated Intelligent Machines Research Corp (IMR Corp) in March, 1951. The goal – productizing GISMO‘s breakthrough text scanning capabilities into workable systems for business, government and military partners.
Within months, early sales confirmed dramatic interest…
GISMO Unit Orders – 1951
| Customer | Units Ordered |
|---|---|
| First National Bank | 3 |
| U.S Defense Dept. | 2 |
| Reader‘s Digest | 1 |
| Eastman Kodak | 1 |
With early-adopter validation in place, Shepard proceeded to enhance GISMO‘s capabilities. A smarter scanning technology and purpose-built font would unlock greater speed and accuracy – vital for real-world usage.
Streamlined Invocation – The Farrington Font
While GISMO impressed technically, Shepard knew specialized fonts could strengthen character recognition. Every typewriter and printing press of the era used customized fonts and styling that challenged scanning reliability.
In 1955, Shepard devised a streamlined, numeric-only font permitting faster and more accurate machine reading. Unadorned by superfluous strokes or ornamentation, this "Farrington Font" increased letter recognition to over 99% accuracy. Equally important – it became the industry standard for early Magnetic Ink Character Recognition (MICR) check printing applications.
When IMR Corp was acquired by MICR firm Farrington Manufacturing in 1959, Shepard‘s purpose-built digital font formed the basis for check and financial processing for decades hence. Its progeny remain in modern credit card and check stock numbering – a lasting testament to Shepard‘s technical prowess.
Lasting Impact
While GISMO itself faded into history, its breakthrough pivot into optical text recognition unleashed a torrent of innovation. IBM and Siemens introduced mainframe OCR systems in the 1960‘s based largely on GISMO‘s approach. The digital scanning techniques Shepard pioneered eventually enabled ubiquitous modern technologies like:
- Automated document analysis
- Form processing software
- Text-to-speech conversion
- Optical character verification
- Signature recognition
- CAPTCHA human detection tests
David Shepard continued prolifically inventing until his passing in 2000, averaging one new patent yearly for 5 straight decades. Yet his first greatadvance – GISMO – remains a milestone in computing history. Its 1952 debut proved that machine identification of printed language was indeed possible – opening vast new vistas for processing, translating and unlocking value from the written word.
