The aging population is increasing the demand for remote long-term care solutions that enable the elderly to live independently at home for longer.
Telekom Slovenije offers two distinct platforms: E-care for remotely assisting the elderly and NEO for multimedia entertainment.
These platforms operate separately, missing the potential for integration.
This master's thesis addresses this gap by developing and evaluating a virtual assistant that combines both platforms into a comprehensive solution to support the elderly.
We developed a prototype system centered around the open-source Home Assistant platform.
This system integrates data from E-care sensors, telemetry from the NEO Smartbox, and smart home devices.
For user interaction, we developed a dedicated Android application, enabling voice control in the Slovenian language.
The system leverages the OpenAI ChatGPT large language model for natural language understanding and the LucamiRecSys recommender system to provide advice on daily activities.
The system's functionality was evaluated through user testing, employing standardized UEQ and UMUX questionnaires and a semi-structured interview.
The results indicated a good user experience.
The system achieved above-average scores for attractiveness, perspicuity, efficiency, and novelty.
Users praised the ease of use, speed, responsiveness, and understanding of the Slovenian language.
They were frustrated by unreliable keyword detection and a limited number of recommendations.
The main contribution of this work is a practical proof-of-concept demonstrating that by integrating existing commercial and open-source technologies, it is possible to develop an advanced virtual assistant for technology-assisted elderly care using the Slovenian language.
|
