The paper “Toward GenAI-Based Contextual Metadata Generation for Reusability: Evaluating Metadata Extraction from Scientific Publications” won second place in the “Young Scientist Award” category at this year’s I4CS conference in Amsterdam.
Congratulations, Burak Toptas!
The study examines how human and LLM-based annotations of contextual metadata from scientific publications differ in terms of completeness and semantic correctness. To this end, a text-based evaluation method was developed that does not treat human annotations as ground truth, and it shows that LLMs often capture more relevant contextual information without systematically generating unsubstantiated content, such as hallucinations.