Weighted finite automata with output (Q1797782)

Several non-classical variants of weighted finite transducers over semirings are introduced and studied in the article. Their definitions are mostly obtained by generalisation of various classes of fuzzy transducers studied in the literature -- see for instance [\textit{Y. Li} and \textit{W. Pedrycz}, Soft Comput. 10, No. 10, 953--959 (2006; Zbl 1109.68059)]. In particular, the authors consider what they call sequential weighted automata with output, three different semantics for a model of weighted transducers inspired by Mealy machines and two semantics for a model inspired by Moore machines. Sequential weighted automata considered in this article should not be confused with sequential automata and transducers as understood, e.g., in [\textit{M. Mohri}, \textit{F. Pereira} and \textit{M. Riley}, ``Weighted finite-state transducers in speech recognition'', Comput. Speech Lang. 16, No. 1, 69--88 (2002)] or in [\textit{S. Lombardy} and \textit{J. Sakarovitch}, Theor. Comput. Sci. 356, No. 1--2, 224--244 (2006; Zbl 1160.68419)] (in particular, no form of input-determinism is required). The main results proved in the article show that each two of these models are equivalent under at least one of the semantics considered.