• DocumentCode
    3432494
  • Title

    New evaluation framework for metadata mapping approaches based on Markov models

  • Author

    Ben Messaoud, Ines ; Amiri, Hamid ; El Abed, Haikal ; Märgner, Volker

  • Author_Institution
    Lab. des Syst. et Traitement de Signal, Ecole Nat. d´´Ing. de Tunis, Tunis, Tunisia
  • fYear
    2012
  • fDate
    2-5 July 2012
  • Firstpage
    764
  • Lastpage
    769
  • Abstract
    Document annotation is considered as the definition of logical and physical structures of document. Several works for page layout analysis have been presented but there is no standard annotation for one document. In order to draw the relationship between these different presentations, a Mapping should be applied between different annotations. The question that should be answered is how efficient is this Mapping? We present in this paper a framework allowing the generation of a standard annotation for each document database. The description of the page layout analysis is based on XML metadata. The Mapping is applied between the standard document and a new document in order to allow the correspondence between them. The Mapping is evaluated using a proposed model based on Markov models. The results show that the probability rate for the Mapping evaluation varies from 0.6 to 1.
  • Keywords
    least squares approximations; mathematical analysis; nonlinear systems; spectral analysis; support vector machines; LS-SVM model; SSA; de-noised data; identification approach; least squares support vector machines; mathematical models; measurement data de-noising; nonlinear system identification; signal processing; singular spectrum analysis; Feature extraction; Hidden Markov models; Layout; Markov processes; Mathematical model; Probability; XML;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Information Science, Signal Processing and their Applications (ISSPA), 2012 11th International Conference on
  • Conference_Location
    Montreal, QC
  • Print_ISBN
    978-1-4673-0381-1
  • Electronic_ISBN
    978-1-4673-0380-4
  • Type

    conf

  • DOI
    10.1109/ISSPA.2012.6310655
  • Filename
    6310655