Deciphering cuneiform texts

Translating the content of tens of thousands clay tablets is a lengthy and time consuming process that can only be compared with the huge amount of papyri scraps that were recovered at Oxyrhynchus. As far as the papyri are concerned, a new computer based program has been put into place (see: Get involved with Oxyrhynchus) to help and accelerate the translations.

Reading cuneiform signs is a far more complicated operation but it is heartwarming to learn that scientists have been able to get extra help from their computer technology as well. The main problem in processing these tablets is the fact that the signs are tri-dimensional and that the cuneiform characters are very diverse.

Thanks to a breakthrough at the University of Chicago, they are now able to make automated transcriptions, especially of those texts that were recovered from the Persepolis area in 1933. Our knowledge about the Achaemenid history is growing rapidly.

This new technology enables to create a database management platform and over the last five years the system has improved significantly. As a result, scholars have created a good machine learning procedure. Currently over 60 terabytes of digitized high-resolution images have been created and are being made ready to return to Iran

The present collection has led to the creation a dictionary of the Elamite language that can be used by today’s students. The machine learning model can successfully decipher cuneiform signs with an accuracy of approx. 80% already. Being able to translate and identify most of the repetitive parts, frees a lot of time for those experts whose task remains to analyze and interpret the difficult place names or signs that still need closer study.

In the future, the deciphering system can be shared with other archaeologists to retrain it in such a way that cuneiform languages other than Elamite can be translated as well. There still is so much to be discovered!