Towards the Processing of Historic Documents

This chapter describes methods required for transforming complex document images into texts. The goal is to make the contents of those documents available for search engines, which are not born-digital but converted from a physical medium to a digital format. Established optical character recognition methods fail for documents for which no assumptions can be made regarding the, probably unknown, symbols contained in the document, historic documents being the example domain par excellence. This paper, however, has a much broader goal: it outlines fundamental problems as well as a methodology in the dealing with documents containing unknown and arbitrary symbols in order to provide a basis for discussions and future work within the digital library community. In particular, future advances will more closely require the interaction of researchers concerned with such diverse topics as document digitisation, reproduction, and preservation as well as search engines, cross-language processing, mobile libraries, and many further areas. Adopting a general view on the presented issues, researchers of the aforementioned areas should be sensitised for the problems met in processing complex, especially historic documents.