Unsupervised Newspaper Segmentation Using Language Context
R. Furmaniak · Proceedings of the International Conference on Document Analysis and Recognition · 2007
There has been increased interest in digitization of news- paper archives. A major problem that must be solved is that of high accuracy decomposition of the page into its logical structure. In this paper I present an approach that uses a language similarity measure based on OCR results to train geometric layout rules tailored to an arbitrary title. Exper- iments have shown this approach to be very effective.