Abstract

This thesis focuses on further developing the KIBeKodA platform, designed to improve law professional’s workflow when analyzing employment contracts. The primary objectives were to implement a versioning system for examination guidelines to enhance the transparency in clause classification and establish intuitive analytics so that lawyers can improve the systems without technical knowledge. Starting with a comprehensive literature review, we provide an overview on how the KIBeKodA system works and how different types of Slowly Changing Dimensions (SCD) are implemented according to best practices. Further, we investigate different metrics that are possibly applicable and help the law professionals to maintain the system themselves. The literature review serves as a basis for the implementations and as a guideline for the successful implementation of the desired functionality. A questionnaire was also developed to assess the system’s usability and intuitiveness in future studies. The results demonstrate that the enhanced KIBeKodA system provides a user-friendly interface and enables legal professionals to optimize their examination guidelines, improving classification outcomes. The prepared questionnaire offers a practical tool for unmoderated usability studies to further refine the system. However, certain limitations were identified, including overly lenient classifications and the need to refine specific metrics further. Despite these difficulties, the system successfully addressed the initial lack of transparency and provided a solid foundation for future enhance- ments. The knowledge acquired paves the way for further development of the KIBeKodA system, guaranteeing that it will continue to be a useful tool for attorneys.

Research Questions

1. How can changes in examination guidelines and their impact on the evaluation of contract clauses in an in-context learning AI system be tracked and assessed?

2. What is needed to create a user-friendly interface which allows to monitor this evolution without technical knowledge?

3. How can technical aspects and the usability of the system be evaluated?