This paper deals with research of patterns and structures in a sequence of data with a finite number of states. After preliminary estimates, two optimization criteria are presented, which based on a schema of referential probabilities enable the construction of a vocabulary of patterns, or basic morphological units. The determination of structures, which may link these patterns is subsequently discussed, first from a morphological viewpoint, then from a generative viewpoint. Morphological structures may be constructed from classical schema of Markov chains or from graph theory; the identification of generative structures is based on the definition of a set of normed transformations on the vocabulary, which leads to various numerical estimates.
CitationDownload as .RIS
MCB UP Ltd
Copyright © 1974, MCB UP Limited