|EUROPA: Research Information Centre
Last Update: 2014-04-16 Source: Star Projects
|View this page online at: http://ec.europa.eu/research/infocentre/article_en.cfm?artid=31957|
Digitising the past
Although millions of books are scanned and put online every year, making old documents and texts available on the web is a difficult and painstaking process.
Project IMPACT which stands for Improving Access to Text - is focused on the making the process easier.
Project IMPACT director Hildelies Balk explained: The problem with turning an historic document into a machine readable text is that it is so very old, everything is different from a modern document, it has old fonts, old words and a very difficult layout.
Once scanned they are left full of errors, because computers struggle to read old texts with strange layouts, fonts and spellings.
Clemens Neudecker, technical manager for European projects at Koninklijke Bibliotheek, showed us one example: This is the Principia Mathematica by Isaac Newton. You see actually what we call shine through, that is ink from the opposite page which is just shining through the paper, you see that the paper is warped, and you can also see here there is this long s also in use, which can very easily be confused with an f.
Researchers at the National Library of the Netherlands have spent four years in a European project to improve software tools to read old books.
Researcher Hildelies Balk said: We improved software for image enhancement, optical character recognition, post-correction of the document and language technology to make it more accessible.
That know-how has already been integrated into the market-leader digitisation software and the results are much improved.
Clemens Neudecker talked us through one project: Here we have an example of the image being straightened. And the next thing is that these borders also need to be cropped. The next step is to transform that into a black and white image in order to enhance the contrast background and foreground.
At the very end of the process the user gets the recognised full text, and theres also the structural features of this text - for example paragraphs, headlines and the like are also detected.
The project claims at least a 15 percent improvement in the accuracy of scanned text.
It means precious archives should be much more available.
Hildelies Balk concluded: Text that is not fully digital, it is virtually invisible. Everyone is used to going into a search engine, and looking for a word, and if they dont find this it basically isnt there for them.