TY - GEN
T1 - Retrieval augmented generation (RAG)-based personalized feedback for construction worker well-being
AU - Gautam, Yogesh
AU - Jebelli, Houtan
N1 - Publisher Copyright:
© 2024 ASCE.
PY - 2024
Y1 - 2024
N2 - Recent strides in wearable sensor technology and artificial intelligence have opened unprecedented avenues for monitoring the health of construction workers on-site. Despite these technological advancements, a critical challenge persists in efficiently communicating real-Time health information to workers. This study bridges this gap by introducing a natural language-based system tailored for providing personalized health status feedback to on-site construction workers. Employing a generative pre-Trained transformer (GPT) model integrated into a retrieval augmented generation (RAG) pipeline, the system crafts individualized messages by integrating data from wearable sensors, machine learning models, and regulatory guidelines. Consequently, the system delivers confidential, context-sensitive feedback to construction workers on-site along with corresponding risk mitigation strategies. The accuracy of the system was evaluated using various performance metrics, achieving a BLEU score of 0.75, a BERT score of 0.3, and a sentence embedding score of 0.9. The findings demonstrated promising results, providing workers with accurate and effective feedback regarding their health status. This innovative communication approach not only addresses a critical research gap but also establishes a new benchmark for personalized feedback systems within the construction industry. Ultimately, this advancement can contribute to enhancing worker awareness and fostering a safer and more inclusive construction work environment.
AB - Recent strides in wearable sensor technology and artificial intelligence have opened unprecedented avenues for monitoring the health of construction workers on-site. Despite these technological advancements, a critical challenge persists in efficiently communicating real-Time health information to workers. This study bridges this gap by introducing a natural language-based system tailored for providing personalized health status feedback to on-site construction workers. Employing a generative pre-Trained transformer (GPT) model integrated into a retrieval augmented generation (RAG) pipeline, the system crafts individualized messages by integrating data from wearable sensors, machine learning models, and regulatory guidelines. Consequently, the system delivers confidential, context-sensitive feedback to construction workers on-site along with corresponding risk mitigation strategies. The accuracy of the system was evaluated using various performance metrics, achieving a BLEU score of 0.75, a BERT score of 0.3, and a sentence embedding score of 0.9. The findings demonstrated promising results, providing workers with accurate and effective feedback regarding their health status. This innovative communication approach not only addresses a critical research gap but also establishes a new benchmark for personalized feedback systems within the construction industry. Ultimately, this advancement can contribute to enhancing worker awareness and fostering a safer and more inclusive construction work environment.
UR - https://www.scopus.com/pages/publications/105025152265
UR - https://www.scopus.com/pages/publications/105025152265#tab=citedBy
U2 - 10.1061/9780784486139.060
DO - 10.1061/9780784486139.060
M3 - Conference contribution
AN - SCOPUS:105025152265
T3 - Computing in Civil Engineering 2024: Sustainability, Resilience, Safety, and Education - Selected papers from the ASCE International Conference on Computing in Civil Engineering 2024
SP - 559
EP - 567
BT - Computing in Civil Engineering 2024
A2 - Akinci, Burcu
A2 - Berges, Mario
A2 - Jazizadeh, Farrokh
A2 - Menassa, Carol C.
A2 - Yeoh, Justin
PB - American Society of Civil Engineers
T2 - 2024 ASCE International Conference on Computing in Civil Engineering, i3CE 2024
Y2 - 28 July 2024 through 31 July 2024
ER -