Term of Award
Master of Science in Mathematics (M.S.)
Document Type and Release Option
Thesis (restricted to Georgia Southern)
Copyright Statement / License for Reuse
This work is licensed under a Creative Commons Attribution 4.0 License.
Department of Mathematical Sciences
Committee Member 1
Committee Member 2
Restoring damaged historical manuscripts and making them available to the large public has been of great interest for humanities researchers long before computers provided assistance for this task. Current technologies and models make this process easier, more accurate, and capable of discovering parts that were previously unknown. I use Recurrent Neural Networks for uncovering hidden Markov models in sequences of characters from historic manuscripts. Such manuscripts are typically written in some archaic language, which makes the underlying machine learning problem inherently difficult, as not much training data is available, in general. I use bidirectional, hierarchical models for sequences of one or more characters, trained on the existent manuscript data. I tested my model and present experimental results using an Old English manuscript.
Khan, Sushmita, "Mining Hidden Markov Models in Sequences of Characters Using Recurrent Neural Networks" (2020). Electronic Theses and Dissertations. 2154.
Research Data and Supplementary Material
Available for download on Monday, April 01, 2030