A Robot That Knows How To Read

Teaching a robot to recognize symbols is one thing; teaching a robot to actually read and understand those symbols is another thing entirely, but researchers in the UK are getting the job done, thanks to a little artificial intuition.

People use text every day. We find our way using street signs, order from written menus, and navigate our way around the mall using large glowing letters as our guide.

Robots can use text as well, but they are generally limited to a more controlled sort of text. For instance, the text on this webpage is generally uniform, with very few variations that would throw off a robotic scanning mechanism. The problem is that text doesn't work like that in real world. Signs use different fonts at different sizes, spacing, and alignments. Even the way light plays off the words in a sign can make it difficult for a robot to discern.

In order to create a robot that could read nearly as well as a human, Ingmar Posner and Paul Newman at the University of Oxford and collaborator Peter Corke at Queensland University of Technology had to address what made humans able to read text better than robots.

So what makes us better?

To a literate human, reading is a simple matter. If a word changes size, or the lighting in the room changes, you don't instantly become illiterate. That's because the human brain can make intuitive leaps of logic, said Edward Grant, Director of the Center for Robotics and Intelligent Machines at North Carolina State University.

"The brain says, 'I've done something kinda like this before, so I can adapt to this new activity that has been presented to me,'" said Grant.

Robots generally cannot make such leaps in intuition. If the font changes, they have to learn to read it all over again. If the word angle changes, same thing.

So how can Marge, the robot created by Posner, Newman, and Corke, find and correct misspelled words, read the BBC Online and the New York Times, and tell a restaurant from a bank?

The answer is actually pretty simple. Her creators loaded her up with Optical Character Recognition software, a spell-checker, and a dictionary. Along with some software tricks to make sure she is reading text and not random cracks in the sidewalk, Marge is able to determine what a word says and what it means, effectively making an artificial intuitive leap.

"On the one hand it makes perfect sense," said Gregory Dudek, a scientist at the Centre for Intelligent Machines at McGill University. "The environment has markings for people to use, and this exploits those human markings for use in a robot."

Why care if robots can read? Aside from the eventual robotic drivers and airplane pilots, this advanced robotic text recognition could be used in devices like glasses that scan the environment and deliver instant information on landmarks, stores, and restaurants. Police cars could instantly scan automobile license plates and alert the patrolman if nearby cars have been reported stolen or have lapsed car insurance.

Okay, maybe that last one is bad.

Robot Can Read, Learn Like a Human [Discovery News]