Thursday, June 28, 2012

#ImpactOCR - Citing a #newspaper

#Wikipedia has this rule: "Citation needed". Much of the news is first published in newspapers. When a citation is needed about something that happened, it is in the newspaper where you will find it mentioned and there may be many chronological entries on the same subject describing how something evolves.

A lot of research and development has gone in the optical character reading of newspaper of the Impact project. As this project has ended and has evolved into a competence centre, its last conference was very much a presentation of what the project achieved.

From my perspective, it produced a lot of software much of it open sourced and all of it is implemented and embedded in the library, archive and research world. It is a world that finds its public for the work done in the Impact project very much in the research world. The general public can benefit as much, what has to be clear is how it could benefit.

Though Europeana newspapers some 10 million newspaper pages will be made accessible. These pages are scanned and to make them really useful they undergo optical character recognition. This is exactly where the Impact project has its impact; as the OCR technology improves, more words are correctly recognised and consequently more content of the newspapers can be discovered.

The results can be improved even further when the public helps train the OCR software recognise characters for specific documents. As citing sources for Wikipedia is an obvious use case for historic newspapers, there are many people who are willing to teach OCR engines to do a better job. For those articles that are found to be particularly useful, proofreading can improve the results even further.

With a public that is involved in improving the digitised and OCR-ed texts everybody will be a winner including the scientific research on these texts.

No comments: