Search papers, labs, and topics across Lattice.
CJM, LIPN
4
0
6
6
VLMs can transform under-resourced historical languages by automating data extraction at unprecedented scales, as demonstrated by the mapping of Armenian commercial advertisements in Paris.
A hybrid method combining semantic detection with LLMs slashes ordering errors by up to 76%, revolutionizing how we approach historical document reconstruction.
A new OCR pipeline slashes error rates on noisy, polytonic Greek texts, opening up a vast historical corpus for NLP research and LLM training.
LLMs can now credibly bootstrap linguistic annotation for historical languages, even without fine-tuning, offering a path around data scarcity.
