Creating Digital Resources from Legacy Documents: An Experience Report from the Biosystematics Domain

Digitized legacy document marked up with XML can be used in many ways, e.g., to generate RDF statements about the world described. A prerequisite for doing so is that the document markup is of sufficient quality. Since fully automated markup-generation methods cannot ensure this, manual corrections and cleaning are indispensable. In this paper, we report on our experiences from a digitization and markup project for a large corpus of legacy documents from the biosystematics domain, with a focus on the use of modern tools. The markup created covers both document structure and semantic details. In contrast to previous markup projects reported on in literature, our corpus consists of large publications that comprise many different semantic units, and the documents contain OCR noise and layout artifacts. A core insight is that digitization and automated markup on the one hand and manual cleaning and correction on the other hand should be tightly interleaved, and that tools supporting this integration yield a significant improvement.