Artwork Capability Assessment of NLP Models (2024)
This project evaluates whether NLP models can understand and reason over art-domain texts. We built a domain-specific dataset by:
- Curating and annotating text from A Text‑Book of the History of Painting and translated Louvre metadata
- Designing a BRAT annotation schema with entities, relations, and events
- Converting annotations to CoNLL for training and evaluation
We fine‑tuned DistilBERT for named entity recognition and performed qualitative inference to assess interpretive capabilities in museum and art‑history contexts.