Development of a dialog system for interaction with robots

• Autor(a) principal: Figueiredo, Sara Cristina Albuquerque
• Data de Publicação: 2014
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/14030
• topic: Engenharia de computadores; Robótica; Robots autónomos; Interacção homem-robô; Processamento da linguagem natural (Ciência de computadores)

Service robots operate in the same environment as humans and perform actions that a human usually performs. These robots must be able to operate autonomously in unknown and dynamic environments, as well as to maneuver with several people and know how to deal with them. By complying with these requirements, they are able to successfully address humans and fulfill their requests whenever they need assistance in a certain task. Natural language communication, including speech that is the most natural way of communication between humans, becomes relevant in the field of Human-Robot Interaction (HRI). By endowing service robots with intuitive spoken interfaces, the specification of the human required tasks is facilitated. However, this is a complicated task to achieve due to the resources involved in creating a sufficiently intuitive spoken interface and because of the difficulty of deploying it in different robots. The main objective of this thesis is the definition, implementation and evaluation of a dialogue system that can be easily integrated into any robotic platform and that functions as a flexible base for the creation of any conversational scenario in the Portuguese language. The system must meet the basic requirements for intuitive and natural communications, namely the characteristics of human-human conversations. A system was developed that functions as a base to give continuity to future work on Spoken Dialog Systems. The system incorporates the client-server architecture, where the client runs on the robot and captures what the user says. The client takes advantage on external dialogue management services. They are executed by the server, which processes the audio obtained, returning an appropriate response given the context of the dialogue. The development was based on a critical analysis of the state of the art in order for the system to be as faithful as possible to what is already done. Through the evaluation phase of the system, it was managed to obtain by few volunteers the conclusion that the main objective was accomplished: a base system was created that is flexible enough to explore different contexts of conversation, such as interacting with children or providing information on a university environment.