AI and ancient languages

Available in English and Arabic, Fabricius is an enjoyable introduction to hieroglyphs, but also presents new approaches for professional Egyptologists.

A new tool, launched this year on the anniversary of the discovery of the Rosetta Stone, offers a novel high-tech way to examine ancient Egyptian hieroglyphs. Fabricius is an open-source program designed by Google Arts & Culture, the Australian Center for Egyptology at Macquarie University, Ubisoft, and Psycle Interactive, in collaboration with Egyptologists from around the world. The digital tool uses Google Cloud’s automated machine learning (AutoML) technology, AutoML Vision, to create a machine-learning model that is able to make sense of what a hieroglyph is. The program includes a variety of features that help users to study and understand Middle Egyptian hieroglyphs – the classical form of the Egyptian language, and the best documented, which replaced Old Egyptian from c.2000 BC.

Available in English and Arabic, Fabricius is an enjoyable introduction to hieroglyphs, but also presents new approaches for professional Egyptologists. The program uses a three-strand approach, offering users the opportunity to learn, play, or work.

The ‘Learn’ tool takes you through the basics of the hieroglyphic script in six steps, each complete with exercises that have you tracing, drawing, identifying, reconstructing, reading, and finally translating hieroglyphs. In addition to introducing some examples of commonly used hieroglyphs, the program explains how hieroglyphic script works and how it has been studied in the past.

Once you have mastered the basics, the ‘Play’ feature allows you to create and share messages in hieroglyphs. You can type your own words and phrases, and select the nearest matches in the ancient Egyptian dictionary, create a message by selecting modern emojis from the list provided, or choose one of the pre-prepared phrases.

The third strand, ‘Work’, is designed for Egyptologists who are already familiar with the process of translating hieroglyphs. The ‘Workbench’ allows users to upload pictures of artefacts bearing hieroglyphs and create more-accurate images of the hieroglyphs, analyse them (quickly identifying hieroglyphs, assisted by the auto-classify function), and translate them using the dictionary service.

Dr Roland Enmarch, a Senior Lecturer in Egyptology at the University of Liverpool, sees Fabricius’ strengths lying in its ability to engage the wider public with issues of recording hieroglyphs. Although there are some difficulties inherent in the application of machine-learning to the recognition of these rather variable hand-created signs, Fabricius offers a ‘fun way for people to learn about some of the issues that occur’. The ‘Learn’ and ‘Play’ features also offer a convenient way for Egyptology students – and anyone else with an interest in the subject – to begin to learn about hieroglyphic epigraphy at home, a resource which, Enmarch observes, is particularly useful in the present COVID-19 pandemic, during which access to the original ancient texts may be restricted.

At present, the project’s aim of increasing the efficiency of the translation of ancient texts appears to be a longer-term goal, according to Enmarch. He highlights that Fabricius is currently focused on the recognition of individual hieroglyphic signs, and while it is developing its ability to identify words made up of multiple signs, the translation of complete sentences from Egyptian into English would require the automated application of syntactic rules, which is not possible just yet. Enmarch also notes that the results of the English to Egyptian translation facility do not directly mimic real hieroglyphic inscriptions. However, the primary goal of the ‘Play’ feature is not to create authentic passages of hieroglyphic script, but to encourage users to experiment and become familiar with hieroglyphs in an entertaining way, and this it undoubtedly achieves.

Fabricius offers an enjoyable introduction to hieroglyphs for beginners and a selection of useful tools to help experts analyse and digitise examples of ancient text, which will doubtless be improved as the program is developed. Perhaps more importantly, though, it represents an important step in the use of technology to transform the way hieroglyphs, and other ancient languages, can be studied.

Further information
Fabricius is available at
IMAGES: Google Arts & Culture, Macquarie University, Psycle Interactive & Ubisoft; courtesy of the Trustees of The British Museum.