In a scene from a video game, a person with a round shield on his back climbs a shiny golden pyramid above plains covered with trees, mountains and a bright yellow sun in the distance.

Can machine learning translate ancient Egyptian texts?

l long intrigued by archaeogaming-an academic discipline that explores the amalgamation of archaeological objects, methods, and characters in video games. So I was really happy when the video game company Ubisoft came out Assassin’s Creed: Originsbet Egypt during the reign of Cleopatra† the designers worked with Egyptologists to ensure that everything from the architecture to the hieroglyphics created an accurate, immersive world. Unexpectedly, this collaboration inspired a machine learning spin-off that changed the course of my early career.

WWhile working with Egyptologists, the game developers learned that translating and interpreting ancient hieroglyphic texts is time-consuming, and that the process has changed little over the past century. So Ubisoft decided to give something back to the academic community through the Hieroglyphics Initiative

tThe project aims to help researchers decode Egyptian hieroglyphs using machine learning — a type of artificial intelligence that allows computer programs to make better predictions through experience.

tThe developers worked with a range of Egyptology scientists, including members of the Australian Center for Egyptology at Macquarie University, where I was a master’s student looking for a topic for a thesis. It felt like a sign. The Hieroglyphics Initiative was an opportunity for me to give back to the academic community by helping to develop a computer program that could benefit Egyptologists. I could explore how machine learning could transform archaeological research and education, and even how the public interacts with Egyptology.

tThe question was: Has the Hieroglyphics Initiative lived up to its promise? I’ve made it my mission to find out.

The difficulties of deciphering hieroglyphs

ddeciphering ancient Egyptian texts has challenged scholars for centuries. Ancient Egyptians used regularly more than 700 hieroglyphs, wrote their lyrics in different directions, avoiding vowels, punctuation, and space between words. Many early researchers mistakenly assumed that hieroglyphs were mere symbols; in fact, they can represent objects, ideas or sound groups.

In a scene from a video game, a person with a round shield on his back climbs a shiny golden pyramid above plains covered with trees, mountains and a bright yellow sun in the distance.

The makers of Assassin’s Creed: Origins consulted with Egyptologists to bring Cleopatra’s Egypt to life. Ubisoft

dDuring the Middle Ages, Arab and Egyptian scholars retained some knowledge of the hieroglyphic system, and European researchers were later informed of their efforts. But scientists in the West did not have a major breakthrough until 1822, when Jean-François Champollion cracked the code by using the multilingual inscription on the Rosetta Stone

AFirst, translators used paper and pencil to sketch or trace inscriptions on tomb walls. Then, in the 1890s, wide-format photography revolutionized the field, allowing scientists to more accurately capture the details of inscriptions and reliefs.

tet Hieroglyphics Initiative seeks to further develop these techniques by applying machine learning to capture and preserve this data. in 2020, Google Arts and Culturewho had taken over the project, released the program in English and Arabic, rename to Fabricius

youUsing this open-source program, researchers can upload a photo of text, such as an inscription on a grave wall, create a facsimile layer of the image, and zoom in on a specific area. They can also use the draw and erase function to touch up areas of the hieroglyphs that may have been damaged by cracks in the rock face.

tif they can arrange the hieroglyphs, and the program can help identify each hieroglyphic character using Gardiner’s Sign List, which groups characters into categories such as birds, reptiles, or human body parts. Based on the sequence and the identified hieroglyphs, the program can then suggest possible translations.

A video shows how Fabricius can clean up hieroglyphs, like these characters from Saroy’s biography. Bree Kelly/Macquarie University

tThe process isn’t as simple or automatic as, say, the camera option on Google Translate, which translates text from images. It would be impossible on its own for a machine to fully integrate all the nuances of ancient Egyptian grammar and spelling with sufficient accuracy to identify hieroglyphic sequences and suggest translations. This is because hieroglyphs can have multiple meanings, their spellings varied over thousands of years, and different scribes and artists had stylistic idiosyncrasies.

ttranslating his resources is complex and involves a high degree of interpretation. However, Fabricius could potentially help Egyptologists start a little further in the analytic process than translating manually. And while Fabricius was initially designed with academics in mind, it is now available to the public, with the aim of making ancient Egyptian writing more accessible to curious non-specialists.

Read more about machine learning: “What if machines could learn like children do?

As part of my research, I made step-by-step instructions for using Fabricius and published articles that provide context to non-experts about the history of decoding hieroglyphics and how egyptologists are implementation of various digital technologies in their research.

for for example, Macquarie University Egyptologist Linda Evans examines depictions of animals in ancient Egyptian art to better understand the environmental history of the area and the relationship between humans and other animals. She analyzed ancient Egyptian murals using DStretch, a plug-in that intensifies and saturates the colors in a digital photo, allowing researchers to see blurred details invisible to the eye. Using this technique, Evans was able to reinterpret several tomb paintings.

tThe work of Evans and others indicates that Egyptologists are interested in innovative new technologies and could potentially benefit from collaborations with companies such as Google and Ubisoft if projects such as Fabricius prove successful.

A digitally altered image on a tomb wall reveals two pigs with bright green and light purple spots surrounding them.

Using software called DStretch, Egyptologist Linda Evans was able to see previously invisible images on a tomb wall. Linda Evans

can machines help translate hieroglyphs?

tTo investigate whether Fabricius is achieving its goals, I launched a case study where I asked eight volunteers with different levels of expertise to test the program. The participants came from Australia, Egypt and the Netherlands, and ranged from Egyptology students to professional scientists. Each participant was given two ancient Egyptian texts to translate – one using their usual translation method, the other using Fabricius.

lat first they were optimistic, but also somewhat skeptical about the ambitious goals of the program. Translation usually involves finding the sounds of certain characters in lists, searching for words in dictionaries and online databases, using grammars to make good translations, and comparing translations of similar texts in various scientific works, both analog and digital. . This complex process is often done on paper and can take several attempts at scribbling, erasing and crossing out symbols.

so, the idea that this multi-step, arduous process could be simplified using machine learning technology and completed on a single computer page intrigued participants.

huhhowever, my research revealed that Fabricius was not as immediately useful as initially hoped. The program was quite behind and the functions were not always obvious to the users. This meant guiding each participant through the process to help them complete a translation in a timely manner. Most importantly, the machine learning models used by Fabricius to suggest characters from tracked photos only achieved a 27 percent accuracy rate.

tThat said, the fact that the program can recognize and organize hieroglyphic characters could make it a useful tool even if the recognition rates aren’t very high. In addition, machine learning allows programs to improve as they receive more and more data. Google Translate, which is powered by a machine learning system, has become more accurate over the years and the performs best in languages it gets more training, like Spanish and German.

The glamor of machine learning may have to be cast aside in favor of Fabriciuspragmatic aspects.

so, since researchers create well-formed training data, a next version of Fabricius could be trained on a much larger dataset. In this way, it could improve accuracy in identifying hieroglyphs over time and help scientists prepare editions of Egyptian texts for use in a digital context, reducing the amount of textual data at their disposal. for text and language research is increasing enormously.

lIn short, Fabricius presents an opportunity to revolutionize the method of recording and conducting research on ancient Egyptian textual data. However, until the program evolves to a point where it can accurately identify at least 99 percent of characters, it’s unlikely to become a tool that complements the manual procedures of text analysis. It could also only be used as an accurate translation tool if the machine learning module could be trained not only with individual characters (the current focus), but also with sentences and larger texts containing grammar and syntax.

lMeanwhile, the glamor and mystique of machine learning may have to be cast aside in favor of the more useful pragmatic aspects of Fabricius’ design, such as the ability to quickly highlight and arrange all hieroglyphs in a text. This procedure alone would help digitize ancient Egyptian hieroglyphic texts much faster so that they can be easily shared, analyzed and uploaded to online databases. Such databases could potentially provide training information for future machine learning applications.

lIn classrooms, Fabricius could generate a new learning method for Egyptology students, encouraging emerging generations to develop skills that could improve the way hieroglyphic data is recorded in an increasingly technological world.

OIn short, Fabricius represents a good first step in building a program that can detect and automatically classify hieroglyphs. It needs to be retrained on a bigger and better dataset before it can be used in a research context. However, the program holds promise as a platform for Egyptologists, students, and laypeople to engage with ancient texts and experiment with hieroglyphic translation. Plus, it could inspire more people to get involved in projects that combine archeology, technology, and video games, like it did for me.

Leave a Comment

Your email address will not be published.