bdd-abbreviations-augmented

The BDD Abbreviations Augmented dataset is a parallel corpus for automated abbreviation expansion in transcriptions from 11th-century Latin manuscripts. It provides paired examples of abbreviated text (source) and its corresponding full expansion (target). The dataset was generated from the five key manuscripts of the Decretum Burchardi, a major work of medieval canon law, transcribed by the Burchards Dekret Digital (BDD) project. It is specifically designed for fine-tuning byte-level sequence-to-sequence models like ByT5 to assist in the creation of digital critical editions. The core task is to translate a "graphemic" transcription, which preserves scribal abbreviations using special Unicode characters (MUFI), into a fully readable, expanded Latin text. Mirror of https://huggingface.co/datasets/mschonhardt/bdd-abbreviations-augmented