Document: a Useful Level for Facing Noisy Data
Hervé Dejean, Jean-Luc Meunier
In this paper we will present a set of experiments using large digitalized collections of books to show that logical structures can be extracted with good quality when working at document level. The proposed solution relies on a twofold method: first specific logical elements are recognized by a given method. Then models for the recognized elements are generated by combining layout, content and labeling information. These inferred models combining several kinds of information are used to correct noisy data, typical zoning, OCR and labeling errors produced by previous processing steps. This method is illustrated with the extraction of page numbers and chapter headings, two navigating elements required by digital libraries.
AND 2010 Fourth Workshop on Analytics for Noisy Unstructured Text Data,
October 26th, 2010, Toronto, Canada
2010-066-Final.pdf (200.47 kB)