Token and Span Classification for Entity Recognition in French Historical Encyclopedias

TL;DR

This paper evaluates various NER methods on 18th-century French encyclopedias, highlighting the effectiveness of transformers and exploring few-shot prompting for low-resource historical texts.

Contribution

It introduces a dual token and span classification framework for nested entities and assesses generative models for low-resource historical NER tasks.

Findings

Transformer models outperform classical methods on nested entities.

Few-shot prompting shows promise in low-resource scenarios.

Hybrid approaches are suggested for complex historical texts.

Abstract

Named Entity Recognition (NER) in historical texts presents unique challenges due to non-standardized language, archaic orthography, and nested or overlapping entities. This study benchmarks a diverse set of NER approaches, ranging from classical Conditional Random Fields (CRFs) and spaCy-based models to transformer-based architectures such as CamemBERT and sequence-labeling models like Flair. Experiments are conducted on the GeoEDdA dataset, a richly annotated corpus derived from 18th-century French encyclopedias. We propose framing NER as both token-level and span-level classification to accommodate complex nested entity structures typical of historical documents. Additionally, we evaluate the emerging potential of few-shot prompting with generative language models for low-resource scenarios. Our results demonstrate that while transformer-based models achieve state-of-the-art performance, especially on nested entities, generative models offer promising alternatives when labeled data are scarce. The study highlights ongoing challenges in historical NER and suggests avenues for hybrid approaches combining symbolic and neural methods to better capture the intricacies of early modern French text.