SUPR
Deep Learning Methods for Historical Language and Cipher Decryption Models
Dnr:

NAISS 2025/22-13

Type:

NAISS Small Compute

Principal Investigator:

Micaella Bruton

Affiliation:

Stockholms universitet

Start Date:

2025-01-15

End Date:

2026-02-01

Primary Classification:

10208: Natural Language Processing

Webpage:

Allocation

Abstract

The aim of the project is to process and decrypt historical encrypted documents and to create tools for (semi)-automatic transcription and decryption. The process of automatic decryption includes hand-written text recognition to automatically convert the manuscript images to machine-readable, transcribed text format and a mapping of symbols with a transcription scheme, the detection of the plaintext language (underlying language) of the cipher on the basis of historical text sources, the automatic identification of the cipher type, the cryptanalysis of the ciphertext, and finally its decryption. I plan to experiment with various types of deep learning algorithms, including zero-shot and few shot learning with and without pretrained language models for the transcription and decryption of ciphers.