Lu Zhang's master thesis passes with distinction!

June 2026.

The master thesis written by Lu Zhang and supervised by UPHG researcher Yunting Xie was awarded the Swedish grade VG (pass with distinction).

The thesis "AI Adoption in Public Institutions: A Case Study on Patent Documents Metadata Extraction"  ".../.../...investigates the use of locally deployed open-source MLLMs for metadata extraction from historical Swedish patent documents. Conducted in collaboration with the Swedish Intellectual Property Office (PRV) and the Patent History Group at Uppsala University, the study combines a technical evaluation of AI-assisted extraction with an examination of AI adoption in public-sector environments. Drawing on a socio-technical perspective and document theory, it explores how technological capabilities interact with institutional requirements and the historical characteristics of patent documents.

An extraction workflow was developed using open-source MLLMs, prompt engineering, and rule-based postprocessing. The workflow was evaluated on historical patent records from 1945–1975. The results demonstrate that locally deployable open-source models can provide a practical alternative to commercial AI services while maintaining data sovereignty and institutional control. The findings further show that successful implementation depends not only on model performance but also on workflow design, document characteristics, and organizational constraints.

The study argues that AI adoption in public institutions should be understood as a socio-technical process rather than a purely technical challenge. It also highlights the limitations of relying solely on overall accuracy as an evaluation metric, demonstrating that different error types have distinct practical implications for large-scale digitization projects. By combining technical evaluation with socio-technical analysis, this thesis contributes to ongoing discussions on AI-assisted database construction, and the responsible adoption of AI in public institutions." (Zhang, 2026, p. 2)

Congratulations to Lu and her supervisor! You can read the thesis here.