EU book digitisation project needs 'Wikipedia'-style army of volunteer editors
By Matej Hruska
An EU partnership with Israeli researchers from US computing firm IBM to digitise major European historical texts is seeking volunteers to help boost the accuracy of scanned texts in a process that will reduce from hours to just minutes the amount of time it currently takes to digitise documents.
The goal of the digitisation project, dubbed Impact (ImProving ACcess to Text), is to increase the accuracy of scanned texts and also editable and searchable online.
Dear EUobserver reader
Subscribe now for unrestricted access to EUobserver.
Sign up for 30 days' free trial, no obligation. Full subscription only 15 € / month or 150 € / year.
- Unlimited access on desktop and mobile
- All premium articles, analysis, commentary and investigations
- EUobserver archives
EUobserver is the only independent news media covering EU affairs in Brussels and all 28 member states.
♡ We value your support.
If you already have an account click here to login.
A new web-based optical character recognition (OCR) technology and online collaboration of institutions aims to help with recognition of texts with faded ink or unusually-shaped typefaces, which are currently scanned only as static images.
The project's researchers believes that the new system will provide between 25 percent and 50 percent greater accuracy than standard recognition programmes.
According to them, the online collaborative correction system hopes to attract volunteers to aid in the process similar to the unpaid editor army that corrects Wikipedia entries and then learn from errors that have been recognised by humans.
The new technology accelerates the process of locating questionable text scans and then enables reviewers to key in corrections to the text. Instead of displaying an entire scanned page, reviewers only see the actual letters or words in question. For example, the letter combination "r" and "n" can sometimes be difficult for a computer to distinguish from the letter 'm'. In these cases, the system collects a variety of instances of the letter 'm' and places these samples next to the letters in question, making it much easier to determine the letter's real identity.
And where an entire word is suspect, it is added to a collection of other questionable terms, which are then arranged in alphabetical order. Volunteer reviewers then just accept or reject suggested substitutes with one keystroke.
Previously, a small book that normally takes four hours to key in manually or one hour using standard OCR technology with manual correction. But the new system cuts the process down to 30 minutes. Researchers believe they will soon be able to cut the time down to 15 minutes as the system enriches its dictionary, learning from the human volunteers.
Brussels and IBM announced on Thursday that they are to expand the new technology to some two dozen national libraries, research institutes, universities and companies, including the British Library, the German National Library and the Poznan Supercomputing and Networkign Centre in Poland.
The European Commission delivered fresh attention to digitisation last year after Google said it planned to make millions of books available on-line, a move that disquieted some European publishers and copyright owners.