Utveckling av ett AI-drivet stödsystem för navigering av administrativa system

Abstract

This thesis focuses on developing an AI-driven support system to improve the navigation of the intranet at Chalmers University of Technology. The aim is to facilitate users’ ability to quickly find relevant information, such as manuals, links to external websites, and contact information.The method included the development and implementation of different AI models to create a system capable of handling both Swedish and English. Several models were tested and compared to identify the most optimal one for the task. The application was evaluated with real data from the economy department of Chalmers, to measure its performance and accuracy. The results exhibit that KBLab/sentence-bert-swedish-cased was the most efficient model, making a correct and precise search possible. The support system displayed its capability to effectively find relevant information from its database based on users’ search terms, thereby improving the user experience and streamlining the workflow. The conclusion is that the developed support system could improve searching for information on Chalmers’ intranet significantly. Future versions of the system could be further developed with improved AI models and training algorithms, as well as with improved security measures and a more intuitive user interface. Thereby, this thesis lays a solid foundation for further development and implementation of advanced support systems within similar organizations. This report is written in Swedish.