Virtual Library of' Multilingual Multicontents Scientific Journals

摘要

The work presented in this paper is tackling the document analysis and recognition problem, in line with a virtual library project. It presents the design and the implementation of a software platform allowing to place at the disposal of end users of a virtual library, automatic processing to display and retrieve, remotely and locally, dematerialized documents from databases made up of multilingual and multicontents scientific journals. The prototype was called BVMuLS. After a chain of digitalization, conversion and compression, the scanned documents are put at the format DjVu and the articles are safeguarded with format PDF. The whole is then stored in a MySQL database. The outline journal is used as index. The contents will be presented at the user as an XML document, allowing a lot of services: consultation, printing and safeguard locally. With this intention, we initially start with the segmentation of the image of the contents in order to extract the area where the text of the contents is. Once the area recognized, we apply a tagging technique so to find the various identifying fields of the articles, namely : the title, the author, the translator and the number of pages. In order to generate the XML file corresponding to this contents, other metadata, such as the name of the review, the volume of the review, the date of publication, etc, are added with those referring to the various articles published in the review.