Finite State Stack Transducer for Functionality in a Multi-Parameter Environment

Objective: This paper introduces and implements Finite-State-Stack-Transducer (FSST), an enhancement of the Finite-State-Transducer (FST) automaton. FSST is an automaton equipped with a stack to maintain the subjective memory suitable for an application. Analysis: Based on the 2-tape FST, the FSST employs this memory stack to select a suitable input alphabet for transition of state according to additional parameters, thus improving its responsiveness to a wider range of inputs. Findings: In this paper, we implement this model in a multi-state environment of a virtual agent to present its mood changes according to type and relevance of input statements. We also prove its utility for modeling conversational virtual agents for interactive environments. Applications: The presented FSST can be employed to improve the emotional reaction of the virtual agents. This improvement can be utilized to enhance the design and response of cognitive agents and automated virtual assistants.