{"id":853,"date":"2023-11-01T13:20:58","date_gmt":"2023-11-01T13:20:58","guid":{"rendered":"https:\/\/todaysainews.com\/index.php\/2023\/11\/01\/predicting-the-past-with-ithaca\/"},"modified":"2025-04-27T07:31:53","modified_gmt":"2025-04-27T07:31:53","slug":"predicting-the-past-with-ithaca","status":"publish","type":"post","link":"https:\/\/todaysainews.com\/index.php\/2023\/11\/01\/predicting-the-past-with-ithaca\/","title":{"rendered":"Predicting the past with Ithaca"},"content":{"rendered":"

[ad_1]
\n<\/p>\n

\n
\n
\n

Research<\/p>\n

\n
Published<\/dt>\n
\n
Authors<\/dt>\n
\n

Yannis Assael, Thea Sommerschield *, Brendan Shillingford, Nando de Freitas (*external authors)<\/p>\n<\/dd>\n<\/dl>\n

\n<\/section><\/div>\n\"\"\n <\/picture>\n<\/p><\/div>\n
\n

Restoring, placing, and dating ancient texts through collaboration between AI and historians<\/b><\/p>\n

The birth of human writing marked the dawn of History<\/a> and is crucial to our understanding of past civilisations and the world we live in today. For example, more than 2,500 years ago, the Greeks began writing on stone, pottery, and metal to document everything from leases and laws to calendars and oracles, giving a detailed insight into the Mediterranean region. Unfortunately, it\u2019s an incomplete record. Many of the surviving inscriptions have been damaged over the centuries or moved from their original location. In addition, modern dating techniques, such as radiocarbon dating<\/a>, cannot be used on these materials, making inscriptions difficult and time-consuming to interpret.<\/p>\n

In line with DeepMind\u2019s mission<\/a> of solving intelligence to advance science and humanity, we collaborated with the Department of Humanities of Ca’ Foscari University of Venice<\/a>, the Classics Faculty of the University of Oxford<\/a>, and the Department of Informatics of the Athens University of Economics and Business<\/a> to explore how machine learning can help historians better interpret these inscriptions \u2013 giving a richer understanding of ancient history and unlocking the potential for cooperation between AI and historians.<\/p>\n

In a paper<\/a> published today in Nature<\/i>, we jointly introduce Ithaca, the first deep neural network that can restore the missing text of damaged inscriptions, identify their original location, and help establish the date they were created. Ithaca is named after the Greek island in Homer\u2019s Odyssey<\/i><\/a> and builds upon and extends Pythia<\/a>, our previous system that focused on textual restoration. Our evaluations show that Ithaca achieves 62% accuracy in restoring damaged texts, 71% accuracy in identifying their original location, and can date texts to within 30 years of their ground-truth date ranges. Historians have already used the tool to reevaluate significant periods in Greek history.<\/p>\n

To make our research widely available to researchers, educators, museum staff and others, we partnered with Google Cloud<\/a> and Google Arts & Culture<\/a> to launch a free interactive version of Ithaca<\/a>. And to aid further research, we have also open sourced<\/a> our code, the pretrained model, and an interactive Colaboratory notebook.<\/p>\n<\/div>\n

\n

Figure 1. This restored inscription (IG I3 4B) records a decree concerning the Acropolis of Athens and dates 485\/4 BCE. (CC BY-SA 3.0, WikiMedia).<\/p>\n<\/figcaption><\/figure>\n

\n

Figure 2. Ithaca’s architecture. Damaged parts of a text are represented with a dash “-“. Here, we artificially corrupted the characters “\u03b4\u03b7\u03bc.” Provided with these inputs, Ithaca restores the text, and identifies the time and place in which the text was written.<\/p>\n<\/figcaption><\/figure>\n

\n

Collaborative tools<\/h2>\n

Ithaca is trained on the largest digital dataset of Greek inscriptions<\/a> from the Packard Humanities Institute<\/a>. Natural language processing<\/a> models are commonly trained using words because the order in which they appear in sentences and the relationships between them provide extra context and meaning. For example, \u201conce upon a time\u201d has more meaning than each character or word seen separately. However, many of the inscriptions historians are interested in analysing with Ithaca are damaged and often missing chunks of text. To ensure our model still works when presented with one of these, we trained it using both words and the individual characters as inputs. The sparse self-attention mechanism at the model\u2019s core evaluates these two inputs in parallel, allowing Ithaca to evaluate inscriptions as needed.<\/p>\n<\/div>\n

\n

Figure 3. Ithaca\u2019s outputs. (a) Restoration predictions for 6 missing characters (dashes) in an Athenian inscription (IG II2 116). The top restoration, in green, is correct (\u03c3\u03c5\u03bc\u03bc\u03b1\u03c7\u03af\u03b1, “alliance”). Note how the following hypotheses (\u1f10\u03ba\u03ba\u03bb\u03b7\u03c3\u03af\u03b1, “assembly” and \u03c0\u03c1\u03bf\u03be\u03b5\u03bd\u03af\u03b1, “treaty between State and foreigner”), highlighted in red, typically occur in Athenian political decrees, revealing Ithaca’s receptivity to context. (b) Geographical attribution of an inscription from Amorgos (IG XII 7, 2). Ithaca’s top prediction is correct, and the closest predictions are neighbouring regions. (c) Date distribution for an inscription from Delos (IG XI 4, 579). The ground-truth date interval 300-250 BCE is in grey; Ithaca’s predicted distribution is in yellow and has a mean at 273 BCE (in green).<\/p>\n<\/figcaption><\/figure>\n

\n

To maximise Ithaca’s value as a research tool, we also created a number of visual aids to ensure Ithaca\u2019s results are easily interpretable by historians:<\/p>\n