Tag
This paper studies unsupervised strategies (TSDAE and unsupervised SimCSE) for adapting specialized language models into corpus-specific sentence encoders for ancient languages, showing they outperform baselines on biblical text-reuse detection in Latin and Ancient Greek, and releases the pipeline as the Paraphrasis online tool.
The paper presents a method for learning word-level explanations of text reuse in Latin literature with minimal annotation, using synthetic training pairs plus active learning; a joint pointer network matches full-data performance with only 200 annotated pairs out of 1,490 references.