Introducing Brain-like Concepts to Embodied Hand-crafted Dialog Management System

The paper introduces a dialog management system that incorporates neurobiological concepts, such as mirror neurons, to create a behavior architecture that blends hand-crafted design with artificial neural networks, aiming to enhance natural language interaction in embodied systems like virtual receptionists.

Abstract

Along with the development of chatbot, language models and speech technologies, there is a growing possibility and interest of creating systems able to interface with humans seamlessly through natural language or directly via speech. In this paper, we want to demonstrate that placing the research on dialog system in the broader context of embodied intelligence allows to introduce concepts taken from neurobiology and neuropsychology to define behavior architecture that reconcile hand-crafted design and artificial neural network and open the gate to future new learning approaches like imitation or learning by instruction. To do so, this paper presents a neural behavior engine that allows creation of mixed initiative dialog and action generation based on hand-crafted models using a graphical language. A demonstration of the usability of such brain-like inspired architecture together with a graphical dialog model is described through a virtual receptionist application running on a semi-public space.

Method

The authors used a methodology that combines neurobiological inspiration, model-driven development (MDD), and a graphical domain-specific language (DSL) to create a dialog management system that integrates hand-crafted models with artificial neural networks. This approach involves the use of "Mirons" to represent linguistic intents and actions, a recurrent neural network-based state machine for dialog management, and a code generation process from the graphical DSL to configure the behavior engine.

Main Finding

The authors discovered that by integrating concepts from neuroscience, specifically mirror neurons, into the design of dialog systems, they could create a behavior architecture that effectively combines hand-crafted models with artificial neural networks. This approach allows for the creation of mixed-initiative dialogs and action generation using a graphical language, which simplifies the design process and overcomes scalability issues. They also found that this methodology could be successfully applied to a virtual receptionist application, which was positively evaluated in terms of usability and attractiveness by the participants of a user study.

Conclusion

The conclusion of the paper is that the integration of neurobiological concepts, such as mirror neurons, into dialog management systems can lead to the development of effective and scalable mixed-initiative dialog systems. The authors demonstrate that their approach, which combines hand-crafted models with artificial neural networks and utilizes a graphical language for dialog design, can be successfully implemented in practical applications like virtual receptionists. The user study conducted on the virtual receptionist application showed positive results, indicating that the system is both usable and attractive to users. The authors suggest that future research should focus on learning methods inspired by infant development, such as imitation learning, instructional learning, or curriculum learning, to further advance the field of dialog systems.

Keywords

dialog management system, neurobiology, mirror neurons, natural language understanding (NLU), natural language generation (NLG), model-driven development (MDD), graphical domain-specific language (DSL), recurrent neural network, state machine, virtual receptionist, usability, attractiveness, user study, embodied systems, mixed-initiative dialogs, behavior architecture, hand-crafted models, artificial neural networks, Mirons, code generation, behavior engine, dialog design, embodiment, multi-modality, speech technologies, voice assistant, chatbot, language models, speech recognition (ASR), natural language processing (NLP), intent recognition, named entities, dialog state, context tracking, speech acts, turn-taking, planning, modularization, parametrization, error handling, facial recognition, proximity detection, telephony, email communication, system architecture, avatar framework, speech synthesis, speech technologies, AttrakDiff scale, System Usability Scale (SUS), user experience, human-machine interaction, embodied intelligence, neuropsychology, behavior architecture, learning approaches, imitation learning, instructional learning, curriculum learning.