Kassegne S., Zemene D. et al / San Diego State University
Ethiopian script (384 characters in the extended version), Armenian (liturgical) (38), Georgian (Asomtavruli) (38) and Caucasian Albanian (52)
The AI did not have access to historical records, religious texts, geographic information or cultural context: it only analyzed the curves, angles, straight lines and general structure of each graphic sign. Armenian letters were, in structural terms, the most similar to Ethiopian
Historians have noted that some letters in the Armenian, Georgian, and Caucasian Albanian alphabets closely resemble Ethiopian charactersthe ancient script of the Horn of Africa.
Until now, most comparisons have depended on the visual observation of letters by academics; however, this method is difficult to verify or reproduce. To overcome this problem, researchers at San Diego State University turned to Artificial intelligence.
No, published last week in the magazine Digital Scholarship in the Humanitiesresearchers trained an AI model with more than 28,000 images of Ethiopian characters.
The model then measured the structural similarity between these characters and letters from three other alphabets developed in the Caucasus region of Eurasia. Among them, the Armenian alphabet was the one that revealed the greatest proximity.
The team designed the model to focus exclusively on the form of each character. TO AI did not have access to historical records, religious texts, geographic information or cultural context. He analyzed only the curves, angles, straight lines and the general structure of each graphic signal.
After being trained with Ethiopic, the model compared these patterns with those of the scripts Armenian, Georgian and Caucasian Albanian, then calculating their similarity using mathematical distance metrics.
The analysis showed that the Armenian letters they were, in structural terms, the most similar to Ethiopian, while Caucasian Albanian and Georgian had less similarity.
As control element, the team included the Latin alphabetwhich revealed a much lower similarity. This suggests that the patterns detected among Caucasus scripts unlikely to be random.
“Our goal was to go beyond visual impressions that are difficult to test or reproduce,” he said. Sam Kassegneprofessor of mechanical engineering at SDSU and principal investigator of the study, cited by .
“By making our criteria explicit and mathematical, we introduce an objective computational approach that is easily reproducible. We believe this reproducibility is the main contribution of our method”, he added.
The conclusions become even more interesting when seen in their historical context. THE Armenian alphabet was created around 405 AD., at a time when Ethiopic was already in use and was spreading throughout the region.
Historical records mention Ethiopian travelers passing through Jerusalem, Egypt and Syria during that period. Mesrop Mashtotswho is credited with inventing the Armenian alphabet, also traveled through the Middle East at that time.
The AI model did not receive any of this information historical. Still, he identified Armenian as the closest structural match to Ethiopic, coinciding with the period when contact between these cultures was most likely.
“What makes the study significant is the fact that the computational geometry and historical research converge in the same writings and in the same period of time”, said Daniel ZemeneSDSU graduate student and first author of the study.
“The model did not have access to historical records, but learned exclusively from visual and structural data and identified Armenian as the closest structural correspondence to Ethiopic precisely within the period that historians have long debated. This convergence between computing and history is remarkable,” he added.
The researchers emphasize that the conclusions have important limitations. The fact that two writings look similar doesn’t mean that one has been copied from the other.
Writing systems may develop similar characteristics independently, or both may havebe influenced by another source similar.
Throughout history, Greek, Roman, Persian, and Arabic scripts have influenced each other in ways that were not always direct or deliberate.
