Carefully, the finger turns the yellowing page of an ancient manuscript. Absorbing the wisdom from centuries past in the reverent silence of the library. A couple of seconds later, another page is turned, then another…
This is no speed-reading student. The finger, in fact, belongs to a remarkably dextrous and efficient robotic book scanner.
The scanner is an nifty solution to an old problem: how to archive and preserve ancient manuscripts in a way that makes them accessible for a modern-day readership.
The robo-scanner has two attractions. First, it takes a careful approach to handling the originals – unlike so-called destructive scanning techniques which involve removing pages from their bindings. Second, it can scan up to 2,500 pages an hour using a unique and fully automatic page-turning system – the robotic ‘bionic finger’.
In contrast, a human manually turning the pages of a book will probably only mange to scan 200 pages an hour, according to the volunteer digitising effort Project Gutenberg.
“The patented bionic finger automatically uses the right pressure to turn pages so as not to damage what it is scanning. The V-shaped cradles limit the pressure put on bindings, while glass plates hold the pages flat as they are scanned,” says Sofie Quidenus, the founder of Quidenus Technologies, the company behind the scanner.
The machines are already at work across the globe, on scanning projects which include ancient Arabic manuscripts for the National Library of Abhu Dhabi, the historical archives of the city of Buenos Aires in Argentina, and helping to digitise the entire collection of the National Library of Norway at a rate of 55,000 pages a week.
The handwritten frontier
Considering that 20 per cent of archive documents are written by hand rather than printed, the next challenge is handwritten text recognition (HTR) technology that converts handwriting into a machine readable format.
“I realised when working on my first start-up that conserving data is not sufficient,” says Quidenus. “Handwriting needs to be digitised so that it is searchable and can be linked to pictures and other types of information.”
Together with Stephan Dorfmeister, she has cofounded SearchInk which will use artificial intelligence techniques such as deep and transfer learning to digitise handwritten documents.
It empowers people to access human knowledge that is locked in handwritten documents
The idea is to teach the algorithm to digest and analyse text in the same way as a person would – to ‘read’ rather than to simply extract and transcribe the text.
Collaborating with other European research centres, the project is initially focused on analysing handwriting in seven languages. The implication for future research is huge – thousands of ancient manuscripts could become searchable and easily accessible.
“What is really exciting about this latest stage for me,” says Quidenus, “is that it empowers people to access human knowledge that is locked in handwritten documents.”
