Optical Character Recognition refers to the extraction of text from an image. It already has applications in passport capture and the mass reading of number plates. OCR’s development began in the late 19th Century, and continues to this very day. The following capture a few memories in the life of OCR so far.
An 18-year-old Mary Jameson demonstrates the Optophone. Light from a printed page reflects on a selenium cell, and the machine lets out a musical chord. Ms. Jameson, who is blind, can read the print at a record-breaking speed of one word per minute. It’s a good start, but the future has better toys in store.
David Shepard develops a machine that could recognise all 26 letters of the Latin alphabet, as produced by a standard type-writer. He calls it “Gismo”, which later evolves into the Farrington Machine. By the 1960s, OCR technology is being used en-masse in mail-sorting by the U.S. postal service.
Kurzweil Technologies releases the CCD flatbed scanner, the first omni-font optical character recognition system. Before 1974, OCR could only read fonts specially designed for machine readability. In the 70s, the scope of OCR is widening, and the technological universe is beginning to imagine its future impact.
The Newton MesssagePad is launched on May 29th, venturing into a complicated realm of OCR technology called handwriting recognition. The Newton is a failure commercially, but OCR is a door slowly coming ajar — and about to be kicked open.
Google takes the OCR software, Tesseract, under its wing, accelerating industry collaboration. In the coming years, OCR is married to neural networks. Instead of having to input the rules of language into a machine, neural networks will allow OCR to recognise patterns for themselves.
The dream of OCR recognising any letter or number, in any language, whether on a nicely-typed printed page or “in the wild”, is only starting to be realised. With its applications quickly expanding, what could Optical Character Recognition recognise next?