ARHIVSKI VJESNIK 41. (ZAGREB, 1998.)

Strana - 70

S. Nikolovska, B. Asprovska, S. Dragićević, O. Iliev, The integrated information system of the Archives of Macedonia, Arh. vjesn., god. 41 (1998), str. 67-86 - Application of artificial neuron networks This method uses models based on artificial neuron networks (Kwok 89), where the search is reviewed as the spreading in activation of an interactive three-level neuron network, while the learning process is based on the back-propagation method. Therefore the joint principles on which these methods are based are: - Documents and questionnaires are reviewed as two different and independent entities. There is no effective link (of any type) between them. - The search is a process that is initiated by a unique questionnaire based on a unique and closed set of documents. The search does not imply interaction between these two categories. - The relevance (which represents a critical parameter in modeling, i.e. whether a document is relevant or not for the defined questionnaire) is reviewed as a third entity that needs to be known a priori. This model can be shown in the following structural scheme: It must be stressed that the application of these methods is connected with previous theoretical and applicative research in order to get the optimal solution for the structural set up of archive materials. 3. INTEGRATED INFORMATION SYSTEM By means of object-oriented analysis of the Archives of Macedonia (which covers its structural, functional and behavioral aspect), the basic directions of development of the informational model of the Archives were derived. The structural model of the Archives is shown in Figure 1. 70