Innovations in machine learning: a case study of the Fabricius Workbench

This thesis investigates how the open-source computer program called the Fabricius Workbench complements the process of translating Hieroglyphic Egyptian texts. The Workbench, developed by Google, Ubisoft, and Psycle Interactive, employs machine learning in an attempt to speed up translation, as has been successfully carried out for other - even ancient - languages. The Workbench utilises machine learning to identify images of Egyptian hieroglyphs. Users can edit a facsimile layer and reconstruct damaged sections of text. The program also suggests words to assist the user in formulating their translation. Workbench project files are stored in a format that is easily shared and edited. By employing eight volunteer Egyptologists of varying skill levels to produce a translation using the Workbench, this thesis evaluated whether there are elements of the program that demonstrate how digital tools might improve the translation process. After analysing the outcomes of the case study by considering the strengths, weaknesses, opportunities, and threats (SWOT) of the program, focusing on user experience, initial expectations of the Workbench had to be reconsidered. The program as it stands would require significant improvements to become a viable tool for Egyptologists. As such, focusing on the individual components of the Workbench would offer earlier rewards and develop a community of users who could demonstrate academic outcomes, thus encouraging further development. Therefore, the program either needs to split into its smaller components or pivot to a tool within a pedagogical program that would sustain the established userbase. Since the Workbench demonstrates that digital technology can be used to capture, manipulate, and analyse hieroglyphic information, it is suggested that students of Hieroglyphic Egyptian could be presented with activities and exercises that show them how to encode and mark up hieroglyphs in order to contribute to the amount of digital textual material of the Ancient Egyptian language available worldwide -- abstract.