Wednesday, December 19, 2018

Translation algorithms are coming for Sumerian!

TECHNOLOGY WATCH: The Key to Cracking Long-Dead Languages? Tablets from some of the world’s oldest civilisations hold rich details about life thousands of years ago, but few people today can read them. New technology is helping to unlock them. (Sophie Hardach, BBC).
[Assyriologist Émilie] Pagé-Perron is coordinating a project to machine translate 69,000 Mesopotamian administrative records from the 21st Century BC. One of the aims is to open up the past to new research.
(Bold emphasis in the original.) There's more to this article, so read it all. But I'm going to focus on this one story.

In the past I have been skeptical about such efforts, which tend to be overblown by the media. (See here, here, and here.) But this one sounds more credible.

Dr. Pagé-Perron's team is training algorithms to translate several thousand cuneiform economic texts and then they intend to let the algorithms loose on the rest. Since economic texts tend to be formulaic and to deal with a limited range of subjects, this could just about work. And once the algorithms have basic competence in economic Sumerian, there's no reason why they can't keep incrementally improving, with humans guiding them initially through progressively more challenging texts.

The algorithms are still quite limited. For example, they need humans to transliterate the cuneiform signs, a very difficult process in itself, and one not entirely disconnected from translation. But there is a big effort ongoing to digitize images of all cuneiform tablets. Once that is done, transliteration of them is a problem that could be attacked with algorithms too.

Not too far in the future, philology may go the way of factory automation.

Cross-file under The Singularity is Near.

Visit PaleoJudaica daily for the latest news on ancient Judaism and the biblical world.