LLMs for Named Entity Recognition in Historical Texts
A new paper on arXiv (2508.18090) explores the use of large language models (LLMs) for Named Entity Recognition (NER) in historical texts. NER identifies proper names like people, organizations, locations, and dates. Traditional supervised methods require large annotated datasets, which are scarce for historical documents due to high labeling costs and expertise needs. Historical language also suffers from inconsistent spelling and archaic vocabulary. The study investigates LLMs' ability to perform NER without extensive training data, addressing these challenges.
Key facts
- arXiv paper 2508.18090
- Focuses on NER for historical texts
- LLMs used as alternative to supervised learning
- Historical texts lack annotated datasets
- Challenges include spelling variability and archaic language
- NER identifies people, organizations, locations, dates
- Supervised approaches require large annotated data
- Paper explores LLM versatility in NLP tasks
Entities
Institutions
- arXiv