Abstract
The Ph.D. thesis investigates the potential of artificial intelligence (AI) in analyzing ancient historical manuscripts, focusing on the Dead Sea Scrolls (DSS) images. The research employs several computer vision, pattern recognition, and machine learning techniques to address writer identification and dating challenges. An initial study highlights the successful application of character shape features, achieving high accuracy in identifying multiple authors within the DSS collection.
After recognizing the crucial role of binarization (extracting ink traces from the background materials) for accurate writer identification, the thesis introduces BiNet, an artificial deep neural network. BiNet utilizes multispectral images and outperforms traditional models in binarizing highly degraded ancient manuscripts, facilitating improved calculation of textural and allographic features. Building upon the success of BiNet, the study identifies multiple authors for the Great Isaiah Scroll, one of the longest scrolls in the DSS collection. The quantitative findings propose a hypothesis contrary to established assumptions about the scroll's authorship.
With the success of writer identification, the thesis employs support vector regression and a self-organizing time map for broader time period classification. Enoch, a Bayesian regression-based model for date prediction, integrates AI with radiocarbon dating, presenting a pioneering technique for estimating manuscript dates.
The interdisciplinary fusion of AI with historical research enhances our understanding of writers' identities, the dating of ancient manuscripts, and the contextualization of historical narratives. The thesis advances methodologies for analyzing ancient manuscripts, contributing to improved interpretations of the past and laying the foundation for further interdisciplinary exploration in historical document analysis.
After recognizing the crucial role of binarization (extracting ink traces from the background materials) for accurate writer identification, the thesis introduces BiNet, an artificial deep neural network. BiNet utilizes multispectral images and outperforms traditional models in binarizing highly degraded ancient manuscripts, facilitating improved calculation of textural and allographic features. Building upon the success of BiNet, the study identifies multiple authors for the Great Isaiah Scroll, one of the longest scrolls in the DSS collection. The quantitative findings propose a hypothesis contrary to established assumptions about the scroll's authorship.
With the success of writer identification, the thesis employs support vector regression and a self-organizing time map for broader time period classification. Enoch, a Bayesian regression-based model for date prediction, integrates AI with radiocarbon dating, presenting a pioneering technique for estimating manuscript dates.
The interdisciplinary fusion of AI with historical research enhances our understanding of writers' identities, the dating of ancient manuscripts, and the contextualization of historical narratives. The thesis advances methodologies for analyzing ancient manuscripts, contributing to improved interpretations of the past and laying the foundation for further interdisciplinary exploration in historical document analysis.
| Original language | English |
|---|---|
| Qualification | Doctor of Philosophy |
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Award date | 23-Jan-2024 |
| Place of Publication | [Groningen] |
| Publisher | |
| Print ISBNs | 978-94-6496-017-4 |
| DOIs | |
| Publication status | Published - 2024 |