Construcao de hierarquia de temas e subtemas de texto

Some Results on Activation and Scaling of Sparse Distributed Memory

Marco GonzalezAna MartinsFernanda BaraoMarceu LeiteVera L. S. de Lima

Jan Kristoferson

This paper presents a contribution to identification of themes and subthemes of Portuguese texts. We build hierarchical structures with terms extracted from a text, trying to define the order of their relative importance and the way how they group. We use techniques, such as stemming, extraction of lexical relations, and frequency weighting; as well as techniques from algorithms and data structures, such as construction of maximum spanning trees. We consider two aspects on the efficiency of Kanerva's Sparse Distributed Memory (SDM). First, it has been suggested that in certain situations it would make sense to use different activation probabilities for writing and reading in SDM. However, here we model such a situation and find that, at least approximately, it is optimal to use the same probabilities for writing and reading. Second, and more important, we investigate the scaling up of SDM, in connection with some observations made by Sjodin. It is shown that the original SDM (here in Jaeckel's version) does not scale up if the reading address is disturbed, but that this can be remedied by using a kind of SDM with sparse address vectors, showing that SDM could well be used as a clean-up memory in computing with large patterns.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: