A recent Google Inc. blog post by Uber Tech Lead Luc Vincent reveals an ex-Hewlett-Packard & Co. employee now working at the search giant has helped to dust off an optical character recognition (OCR) engine with the intent of putting it into the open source domain.
Tesseract, the engine, was developed between 1985 and 1995 by HP Labs, but was tucked away when the company pulled out of the OCR business.
Google called on the Information Science Research Institute at the University of Las Vegas in Nevada, which is known for its expertise in OCR, to help debug the code. With help from researchers at UNLV, the OCR application made its way into the open source domain, Kazem Taghva, the university's associate director for information science, said Wednesday.
Taghva said as an open source project, university researchers and OCR experts could review and improve the application. "One of the guys who once worked for HP now works for Google," he said. "We are working on the project, but Google really has taken the lead in debugging the software."
Today, the Tesseract OCR project, only supports the English language, and does not yet include a page layout analysis module, so it performs poorly on material with multiple columns. "It also doesn't do well on grayscale and color documents, and it's not nearly as accurate as some of the best commercial OCR packages out there," Vincent wrote on the company blog.
With Google's announced plan on Tuesday to provide a service that allows people to search for news articles dating as far back as the 1700s, the reasoning behind the software's resurrection becomes perfectly clear, said David Doermann, associate research scientist at the University of Maryland, College Park. "I'm sure Google intends to automate the process," he said. "They are probably not automatically OCRing them now. Most archives likely have been done by hand."
Doermann said putting the application into an open source project will also help with getting answers to problems not addressed by OCR, such as analysis of complex pages, for example, scanning figures and drawings or text that lay on intricate backgrounds.
It may be too early for Google to take advantage of Tesseract OCR as an open source project to build its digital library, but it could help over the long haul, researchers said.
Through the Google News Archive Search service, Google will work with the New York Times Co. and Washington Post Co., as well as news-retrieval services such as Reed Elsevier Inc.'s LexisNexis, to make articles available.
The Wall Street Journal and Factiva, a news-retrieval service owned by Journal publisher Dow Jones & Co. and Reuters Group plc, also will make articles searchable through the Google service.
Consumers will have an option to search full-text articles using keywords. Google will make summaries available to view for free, but access to the content will require a fee.
Google says it won't host content itself or charge content owners or consumers for the service. Content owners will handle article delivery, pricing and billing to consumers.