The study of ancient manuscripts is a very important field. These manuscripts hold the keys to our past as a human race. Unfortunately, many ancient manuscripts have been degraded to the point where they are no longer legible. Many techniques have been developed to try to read parts of illegible documents. Neural networks are a fairly new technology with an extremely wide range of applications. The goal of this project is to determine how useful a neural network would be as a tool to help in deciphering ancient manuscripts, and, more specifically, Hebrew manuscripts. An image processing application was created to do preprocessing on the characters, and then two neural networks were created to see how well they could perform when analyzing degraded characters. The results were not completely conclusive, but they seem to indicate that a neural network would not be a very good tool to use in analysis of degraded characters.
Hentschel, Daniel, "Creation of a neural network to assist in deciphering degraded ancient Hebrew texts" (1998). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus