Google DeepMind can translate ancient texts thanks to new Aeneas model

artificial robot human ai

In ancient Rome, writing was everywhere, from imperial monuments to everyday objects. But only fragments have survived to this day: many texts have been damaged by time or deliberately destroyed. To help historians reconstruct them, the Google DeepMind team developed Aeneas, an open AI model that can reconstruct Latin inscriptions, determine their age and likely place of origin.

Aeneas works not only with text, but also with context — which is critical in epigraphy. Historians usually look for parallels manually: comparing wording, grammatical constructions and typical phrases to roughly date and interpret a find. Aeneas does this automatically, processing tens of thousands of Latin texts in seconds.

The development was carried out in collaboration with the University of Nottingham with the participation of researchers from Oxford, Warwick and Athens. The model is trained on a specially created dataset LED (Latin Epigraphic Dataset), which includes more than 176,000 inscriptions collected and organized by historians over the decades. The inscriptions are cleaned, unified and provided with metadata: place of discovery, estimated date and context.

Technically, Aeneas uses a transformer architecture. One module analyzes the text, others restore lost symbols and calculate time and geographical parameters. Images indicating the physical origin of the artifact can be used as additional input.

In addition to reconstruction, the model offers a list of the most similar inscriptions – a kind of “contextual affinity”. The system creates a “vector profile” of the text, which takes into account the theme, style, period, origin and similarity with other sources.

The tool is now available online for historians, teachers, students, and museum professionals at predictingthepast.com.

According to the developers, Aeneas shows advanced results and could become a new standard in digital epigraphy, and in the future, adapt to other ancient languages and media, from papyri to coins.


Don't miss interesting news

Subscribe to our channels and read announcements of high-tech news, tes

Leave a Reply

Your email address will not be published. Required fields are marked *





Articles & testsArticles

Oppo A6 Pro smartphone review: ambitious Oppo A6 Pro (CPH2799)

Creating new mid-range smartphones is no easy task. Manufacturers have to balance performance, camera capabilities, displays, and the overall cost impact of each component. How the new Oppo A6 Pro balances these factors is discussed in our review.


In-Ear headphones Active Noise Cancellation: how Oppo eliminate noise Oppo Enco Air5s

In-ear headphones remain one of the most popular form factors due to their lightweight design. However, this design is considered one of the most challenging to implement active noise cancellation (ANC). We’ll explain how Oppo solved this problem.


NewsNews
| 16.02
Poco M8 Power Debuts With Massive 8000mAh Battery and Snapdragon 4 Gen 4

Poco M8 Power is now official, packing a colossal 8000mAh battery, a 120Hz AMOLED display, and the Snapdragon 4 Gen 4, backed by long-term software support.

| 15.52
Generative AI Flunks Misinformation Test: ChatGPT and Gemini Caught Generating Fake News

Researchers at CORRECTIV tested four popular AI tools for their ability to generate fake news graphics. While ChatGPT and Gemini easily bypassed their own safety filters, Meta AI refused to participate in the deception.