🤖 AI Summary
Latin NLP resources remain severely scarce, particularly for medieval legal texts, where high-quality annotated corpora are virtually nonexistent—hindering large language models’ representational capacity on historical legal documents. To address this gap, we introduce LiMe, the first systematically curated corpus of 325 expert-annotated late-medieval Latin criminal verdicts from the Milanese court. Each document is meticulously annotated with part-of-speech (POS) tags, named entities (NER), and syntactic dependencies, and further adapted for masked language modeling (MLM). The corpus integrates high-fidelity digitized texts, structured metadata, and Transformer-optimized pretraining task design. LiMe constitutes the first high-quality, expert-validated resource enabling both supervised learning and self-supervised pretraining for Medieval Latin, thereby establishing critical infrastructure for historical language processing and intelligent analysis of legal paleographic sources.
📝 Abstract
The Latin language has received attention from the computational linguistics research community, which has built, over the years, several valuable resources, ranging from detailed annotated corpora to sophisticated tools for linguistic analysis. With the recent advent of large language models, researchers have also started developing models capable of generating vector representations of Latin texts. The performances of such models remain behind the ones for modern languages, given the disparity in available data. In this paper, we present the LiMe dataset, a corpus of 325 documents extracted from a series of medieval manuscripts called Libri sententiarum potestatis Mediolani, and thoroughly annotated by experts, in order to be employed for masked language model, as well as supervised natural language processing tasks.