A carregar...
Publicação
VASELFCARE - Componente de diálogo inteligente
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 196.11 KB | Adobe PDF |
Autores
Resumo(s)
Esta tese enquadra-se no projeto de investigação VASelfCare cujo objetivo é produzir uma ferramenta informática com um agente relacional, na forma de uma figura humana virtual, que ajuda pessoas mais velhas com diabetes tipo 2 a seguir os seus planos de tratamento. Este assistente virtual ´e um avatar com representação antropomórfica e dotado de inteligência artificial. Este sistema ´e disponibilizado para dispositivos tablet com sistema operativo Android. A aplicação irá ajudar o utilizador em três vertentes: medicação, atividade física e alimentação. O trabalho discutido nesta tese tem como ponto de partida um protótipo já existente, tendo este sido desenvolvido para uma tese de mestrado do ano letivo anterior (2017/18). O objetivo deste trabalho é a incorporação, no assistente virtual, de componentes de inteligência artificial que possibilitem uma interação mais realista entre o assistente e o paciente. Estas componentes incidirão na melhoria do fluxo da interação, com adaptação à situação concreta do doente, considerando todas as dimensões do contexto em causa, bem como na incorporação de componentes que permitam, utilizando informação contextual, enriquecer a dinâmica da atuação do assistente virtual. O desenvolvimento da solução foi seguido da realização de testes com especialistas e utentes em centros de saúde da área de Lisboa que fazem parte do grupo alvo. Para este trabalho em particular foram também realizados testes às diferentes decisões realizadas pela aplicação quando o utilizador ´e ou não aderente a comportamentos adequados em cada uma das componentes trabalhadas na aplicação. Os resultados destes testes foram positivos sendo que as pessoas que testaram a aplicação afirmaram que gostaram de a usar e que gostariam de a usar com frequência. Durante estes testes foram revelados alguns erros que foram resolvidos e foram feitas diversas sugestões que foram ou podem vir a ser implementadas na aplicação.
This thesis fits into the VASelfCare research project whose goal is to produce a computer tool with a relational agent in the form of a virtual human figure that helps older people with type 2 diabetes follow their treatment plans. This virtual assistant is an anthropomorphic avatar with artificial intelligence. This system is available for tablet devices with Android operating system. The application will help the user in three aspects: medication, physical activity and diet. The work discussed in this thesis has as its starting point an already existing prototype, which was developed for a master thesis of the previous school year (2017/18). The goal of this work is the incorporation, in the virtual assistant, of artificial intelligence components that allow a more realistic interaction between the assistant and the patient. These components will focus on improving the flow of interaction, adapting to the patient’s specific situation, considering all the dimensions of the context in question, as well as incorporating components that allow, using contextual information, to enrich the dynamics of the virtual assistant’s performance. The development of the solution was followed by tests with specialists and users in health centers in the Lisbon area that are part of the target group. For this particular work, tests were also performed to the different decisions made by the application when the user is or not adhering to appropriate behaviors in each of the components worked in the application. The results of these tests were positive and the people who tested the app stated that they enjoyed using it and would like to use it frequently. During these tests some errors were revealed that were solved and several suggestions were made that were or may be implemented in the application.
This thesis fits into the VASelfCare research project whose goal is to produce a computer tool with a relational agent in the form of a virtual human figure that helps older people with type 2 diabetes follow their treatment plans. This virtual assistant is an anthropomorphic avatar with artificial intelligence. This system is available for tablet devices with Android operating system. The application will help the user in three aspects: medication, physical activity and diet. The work discussed in this thesis has as its starting point an already existing prototype, which was developed for a master thesis of the previous school year (2017/18). The goal of this work is the incorporation, in the virtual assistant, of artificial intelligence components that allow a more realistic interaction between the assistant and the patient. These components will focus on improving the flow of interaction, adapting to the patient’s specific situation, considering all the dimensions of the context in question, as well as incorporating components that allow, using contextual information, to enrich the dynamics of the virtual assistant’s performance. The development of the solution was followed by tests with specialists and users in health centers in the Lisbon area that are part of the target group. For this particular work, tests were also performed to the different decisions made by the application when the user is or not adhering to appropriate behaviors in each of the components worked in the application. The results of these tests were positive and the people who tested the app stated that they enjoyed using it and would like to use it frequently. During these tests some errors were revealed that were solved and several suggestions were made that were or may be implemented in the application.
Descrição
Trabalho de projecto de mestrado, Engenharia Informática (Engenharia de Software) Universidade de Lisboa, Faculdade de Ciências, 2019
Palavras-chave
Autocuidado Avatar ECA (Embodied Conversational Agent) Inteligência artificial Diabetes Pessoas mais velhas VASelfCare Trabalhos de projecto de mestrado - 2019