An On-Ramp to translation
Users of OCR might be surprised to learn that one of the initial and biggest drivers for the technology has yet to be fully actualized. It was believed soon after the invention of optical character recognition by Ray Kurzweil, that the greatest use of the technology would be in assisting language translation. Even Kurzweil himself very quickly used OCR technology to simply convert scanned image to text so that it could be read digitally for the blind. Some of the developers of OCR technology did not even start with any specialty in imaging but actually specialized in language and dictionary software.
The relationship of OCR technology to language is very interesting and several levels deep. For example, the modern engines show greatest improvements in accuracy by deploying more statistical language models and dictionaries vs. core recognition algorithms. In this method, language is improving the accuracy of OCR technology. For example the letter “e” in English is more frequent than the letter “c”, so in the case where there is a question between an “e” and “c”, this information is useful.
But the most sought after initial use of OCR was simply to get digital text in order to convert it to another language. The dream was to enable travelers to take pictures of foreign signs or documents and have them converted on the fly to their native language. While this was one of the biggest drivers for the further development of OCR, the roadblocks of photography, accurate language translation, and poor processing power of mobile devices was overlooked. Because of this, the use of OCR primarily became document automation and a means to reduce the cost of data entry. This focus changed the way the engines were developed with the new focus being document OCR and not photographic.
I’m confident that the dream will eventually be actualized but I also suspect that many changes to the way OCR engines operate, and the appearance of new specialized engines will happen first.
Chris Riley – Industry Expert