Reconstructing Missing Letters in the Bullinger Correspondence thru Large Language Models

Supervisor(s): Prof. Dr. Martin Volk; Tannon Kew

In this project we aim to reconstruct missing letters in the Bullinger correspondence (ie. letters that have been lost in letter exchanges between certain regular authors). We will use large language models to generate these letters based on hte previous and the next letter in the sequence. This project tries to break new ground by using conditional text generation for a frequent problem in historical document collections. Preliminary experiments lead to nice results.

• Solid background in machine learning
• Python
• Basic knowledge of Latin and German.