Logo des Repositoriums

Auflistung nach Autor:in "Vinay, Rasita"

1 - 2 von 2
  • Konferenzbeitrag
    Demonstrating GRACE: Our Embodied Voice Assistant Providing Cognitive Interventions
    (Mensch und Computer 2024 - Workshopband, 2024) Vinay, Rasita; Tommila, Nora C.; Schlögl, Mathias; Klöppel, Stefan; Biller-Andorno, Nikola; Kowatsch, Tobias
    Our demo describes the development and interaction of our first prototype of GRACE, an embodied voice assistant. GRACE was designed to provide voice-based cognitive interventions to its users, and was first piloted with healthy adults in its current version. The body of GRACE, which was 3D printed, encases the internal components such as the Raspberry Pi, reSpeaker microphone, an AMOLED screen, and a bluetooth speaker. We utilized an open-source robotics platform and their simulation software for executing the script and commands for the voice interactions. A text-to-speech voice from ElevenLabs was used as the voice for GRACE. We were able to develop and design four activities, including two cognitive interactions based on cognitive stimulation therapy. The interaction consisted of an introductory warm-up activity, two cognitive interventions, and a concluding activity.
  • Konferenzbeitrag
    GRACE: Towards an Embodied Voice Assistant for Improving Quality of Life by Leveraging Elements of Cognitive Stimulation Therapy
    (Proceedings of Mensch und Computer 2024, 2024) Vinay, Rasita; Tommila, Nora Camilla; Schlögl, Mathias; Klöppel, Stefan; Biller-Andorno, Nikola; Kowatsch, Tobias
    Dementia is one the leading causes of death globally. Although currently no cure exists, there are multiple interventions aimed at improving the quality of life of people with dementia by leveraging elements of cognitive stimulation therapy. This work is the first of a several-year research agenda that investigates the feasibility and effectiveness of embodied voice assistants to deliver such interventions. We describe the development of the GRACE prototype based on an open-source social robotics platform. We expand its functionality by adding two intervention components of cognitive stimulation therapy and making the robot more interactive by integrating a public large language model. We also conduct a first evaluation with 11 healthy participants who assess perceived technology characteristics including working alliance, an important relationship quality, which is robustly linked to treatment success. Results indicate that GRACE was perceived positively. We conclude this short paper with suggestions for improvement and outline future work.