Parallel Communicating Finite Transducer Systems

Abstract

A parallel communicating finite transducer system is a translating device where several finite transducers work in parallel, in a synchronized manner, and communicate with each other by requests. The communicated data are the current state of the transducer and the current contents of its output tape. Each computation step in such a system is either a usual translating step or a communication step; moreover, the communication steps have priority over the translating ones. Furthermore, whenever a component requests some data, that data must be communicated. We investigate the computational power of these systems. Then we consider systems restricted to subsequential transducers, as components, and compare these systems with the general ones. These devices turned out to be useful in computational linguistics. A short discussion on a possible relevance in the theory of discourse parsing and some directions for further work closes the paper.