Ithaca, the artificial intelligence that can reconstruct incomplete ancient texts

A multidisciplinary team of researchers has created an instrument called Ithaca, an artificial intelligence that analyzes and reconstructs with remarkable accuracy and speed incomplete or damaged ancient texts.

The restoration of ancient texts is one of the main tools that archaeologists use to learn about the past. However, this process is often slow and complicated, as many inscriptions are incomplete or damaged, either because the substrate is a perishable material or because it is broken into several fragments, some of which may be missing.

The most famous case for interpreting ancient texts is that of the Rosetta Stone, but there are many others. Recently, an interdisciplinary group of researchers has developed a tool that uses artificial intelligence to perform the same process much faster and with remarkable accuracy.

The unit in question is called Ithaca, inspired by the island of Ulysses, and operates with the help of a deep neural network: these are computational models based on artificial intelligence with the ability to learn.

When they are given a large database and inputs to analyze (in this case, incomplete text), they calculate a result (a full text proposal) and the percentage probability that it is correct. The full study was published in Nature.

Ancient Greek text restored with Ithaca This inscription (Inscriptiones Graecae, Volume 1, Edition 3, Document 4, Part B (IG I3 4B)) records a decree on the Acropolis of Athens and dates from 485/4 BC Photo: Marsyas , Epigraphic Museum, WikiMedia CC BY 2.5.
Ancient Greek text reconstituted with Ithaca. This inscription is a decree marked on the Acropolis, in Athens, and dates from 485 BC Photo: Marsyas, Epigraphic Museum, WikiMedi

Reconstruction of ancient texts

In the case of Ithaca, tests performed on already deciphered texts showed that the instrument achieves an accuracy of 62% in restoring the complete message, with an accuracy of 71% in identifying the original position of the individual fragments and a margin of error of only 30 years in dating. .

Although they do not seem to be such good percentages, the tool is not meant to work on its own, but to accelerate the work of researchers, especially in the early stages, quickly giving them a reliable hypothesis to work with.

The main purpose for which this tool was developed is precisely to speed up the restoration and assignment of newly discovered texts. Thus, the developers say that they have collaborated with Google to launch a free version of Ithaca, one with a model already “trained” with deciphered texts.

In this way, the tool would be available not only to researchers, but also to other people and institutions that could benefit from it, such as professors and museums.

#Ithaca #artificial #intelligence #reconstruct #incomplete #ancient #texts

Leave a Reply

Your email address will not be published.