Probing Herculaneum Scrolls.

Nicola Davis writes in the Guardian about gingerly attempts to read what’s written in carbonized scrolls:

When Mount Vesuvius erupted in AD79 it destroyed the towns of Pompeii and Herculaneum, their inhabitants and their prized possessions – among them a fine library of scrolls that were carbonised by the searing heat of ash and gas. But scientists say there may still be hope that the fragile documents can once more be read thanks to an innovative approach involving high-energy x-rays and artificial intelligence.

“Although you can see on every flake of papyrus that there is writing, to open it up would require that papyrus to be really limber and flexible – and it is not any more,” said Prof Brent Seales, chair of computer science at the University of Kentucky, who is leading the research. […] Experts have attempted to unroll about half of the scrolls through various methods over the years, although some have been destroyed in the process and experts say unrolling and exposing the writing to the air results in the ink fading. […]

While ink in some Herculaneum fragments has been found to contain lead, Seales says it is only trace amounts and does not allow the inside of the scrolls to be read using x-ray data alone. Seales says it has also proved impossible to replicate findings that letters within Herculaneum scrolls can be deciphered by the naked eye from scans captured by a slightly different x-ray technique. As a result the team have come up with a new approach that uses high-energy x-rays together with a type of artificial intelligence known as machine learning.

The method uses photographs of scroll fragments with writing visible to the naked eye. These are used to teach machine learning algorithms where ink is expected to be in x-ray scans of the same fragments, collected using a number of techniques. The idea is that the system will pick out and learn subtle differences between inked and blank areas in the x-ray scans, such as differences in the structure of papyrus fibres. Once trained on the fragments, it is hoped the system can be used with data from the intact scrolls to reveal the text within. Seales said the team have just finished collecting the x-ray data and are training their algorithms, adding that they will apply the system on the scrolls in the coming months. […]

Dr Dirk Obbink, a papyrologist and classicist at the University of Oxford who has been involved in training the team’s algorithms, said the project was immensely exciting and agreed it is possible the text might turn out to be Latin. “A new historical work by Seneca the Elder was discovered among the unidentified Herculaneum papyri only last year, thus showing what uncontemplated rarities remain to be discovered there,” he said. But Obbink is hoping the scrolls might even contain lost works, such as poems by Sappho or the treatise Mark Antony wrote on his own drunkenness. “I would very much like to be able to read that one,” he said.

I know there’s no actual news here, and I should probably wait until they actually decipher something, but the idea is so exciting (Sappho!) I couldn’t hold off. Thanks, Trevor!


  1. David Eddyshaw says

    said the project was immensely exciting and agreed it is possible the text might turn out to be Latin

    Odd way of putting it. If it’s a text at all, it would surely not be very surprising if it were Latin (though Greek would be more exciting, and Oscan most of all …)

    It’s all worth it for the name Dirk Obbink, though.

  2. David Marjanović says

    Yeah, Oscan has never been found written on papyrus, perhaps just for lack of papyrus…

    I wonder if neutron scanning might help. Make the ink radioactive and see where it radiates.

  3. It’s depressing to think how many have been ruined through ham-handed premature attempts (“hmm, the last dozen we tried just crumbled away, but maybe this one…”).

  4. David Marjanović says

    Probably not that many.

  5. about half

  6. David Marjanović says
  7. This is actually related to an important issue in archeological planning. For important sites, it is common to excavate only a limited part of the site, to preserve some of it for future digs that may have better tools. (The consequences of not doing things that way can obviously be dire. Schliemann dug down to Troy level II in the citadel hill at Hisarlik, finding the ruins of a fortress that was sacked a thousand years before the Trojan War. The center of Troy VI, the one attacked by the Greeks, and many other levels, were permanently lost, leaving subsequent archeologists to digging around the periphery.) Similarly, only a certain fraction of artifacts should be subjected to invasive analyses that could damage or destroy them. (Nobody unwraps and dissects Egyptian mummies any more, but the full autopsies that were done did teach us quite a bit.)

  8. Read the fascinating Guardian article yesterday, and glad to know you too are anxious to know what fragments by which writers might be discovered.

  9. “a type of artificial intelligence known as machine learning”? Oh, the cringe. Is this still how popular publications find it necessary to frame such algorithms?

  10. Lars (the original one) says

    I spotted “… uplifted because of so-called plate tectonic processes …” in Danish Wipe yesterday. (Article on Stevns Klint).

    (Well, it turns out it had been there since 2006 when anonymous pasted in a section that from the style was clearly copied from lower secondary level school material (second form or so). The phrasing is OK in a school book, the kids may never have heard of plate tectonics, but over the years the rest of the section had been updated to encyclopedia style so it stuck out like a sore thumb, as if plate tectonics was a crackpot theory. I fixed it now).

  11. BBC Radio 4’s Inside Science programme went to the lab last week and talked to some of the science team involved. , with the Herculaneum bit starting at 12:06.

  12. D. E. (first comment):
    Hoping for Latin instead of Greek is not odd at all. Most of what has been found at the Villa dei Papiri in Herculaneum has been Greek. Wealthy Romans were generally bilingual and had separate Greek and Latin libraries. The charred rolls found in Piso’s villa at Herculaneum were obviously from Piso’s Greek library. Even worse, they’re almost all prose works of the Epicurean philosopher Philodemus, when most of us would much rather have copies of lost plays of Aeschylus or Sophocles or Ovid’s Medea or Varius’ Thyestes (the two most admired tragedies of the Augustan age) or dozens, if not hundreds, of other lost works.

    Philodemus of Gadara (fellow-citizen of the Gadarene swine) was the pet philosopher of Calpurnius Piso, Julius Caesar’s father-in-law, consul in 58 B.C. and target of Cicero’s hilarious invective In Pisonem. Both were of course long dead by the time Vesuvius erupted, but it appears that Piso’s descendants kept Philodemus’ personal library intact. His three dozen or so mostly-erotic epigrams in the Greek Anthology are excellent (see The Epigrams of Philodemus, ed. David Sider, Oxford, 1997), but the prose is strictly for specialists, being tedious and badly-preserved. There are so many holes in the manuscripts that they can’t really be read as literary works, only interpreted and analyzed, as they have been in dozens of scholarly volumes.

    Classicists on Twitter hear these promises of imminent decipherment of other works several times a year, and many of us are getting tired of the promises. They need to go back to their laboratory and leave us alone until they can show us at least one page, or one paragraph, or even one interesting sentence, of previously-lost literature by someone other than Philodemus.

  13. John Cowan says

    Well, there may be ways of getting a complete copy of Sophokles’ Aleadai, but make sure you don’t take the Romans for granted. They may surprise you. (Search locally on that page for “Death in Vesunna”.)

  14. Dr Dirk Obbink, a papyrologist and classicist at the University of Oxford who has been involved in training the team’s algorithms, said the project was immensely exciting and agreed it is possible the text might turn out to be Latin.

    From today’s paper:

  15. Damn. Why are there so many assholes??

  16. PlasticPaddy says
    This sort of thing happens. The risks are low (until you get greedy or too mixed-up with criminals), the gains are very high (compared to a researcher or librarian’s salary).

  17. Dirk Obbink: the latest news. Isn’t “Christ Church College, Oxford,” in the first sentence, a blatant shibboleth? It’s just “Christ Church,” no “College.” But this is the Guardian and they all went to Oxford. I bet they’re doing it on iconoclastic purpose.

  18. Gotta love the Guardian: “this lugubrious, crumpled, owlish man.” Horrible story, though. I really, really hope the Sappho isn’t a fake.

  19. And yes, Charlotte Higgins, who wrote the story, studied Classics at Balliol College, Oxford. But maybe a subeditor only went to the U of London?

  20. John Cowan says

    And perhaps they just want to explain things to the Gruan’s international audience, who would assume that “Christ Church, Oxford” was a church in the city of Oxford.

  21. They should really put that information in brackets: “Christ Church, Oxford [Note to silly foreign persons: that’s a college, not a church, you ignorant gits].”

  22. John Cowan says

    Dare you say that there are no such ignorant gits in the UK itself? And is it wise to assume that non-UK persons in general know what git means?

  23. But an insult is all the better if the insultee isn’t quite sure what it means. (It’s clear from context that it’s an insult.)

  24. Athel Cornish-Bowden says

    I was going to say that it’s a church as well as a college, but no, it’s got a cathedral within its walls (“Christ Church Cathedral”).

  25. John Cowan says

    I will never really understand the psychology of insult-wielders, as opposed to (say) satirists: “Sticks and stones may break my bones, but names may shame me to the point of suicidal depression.”

  26. Prof. Chris Rollston offers caution about claimed authenticity of “To Ishmaei” paleo-Hebrew fragment:;!!OToaGQ!phnQs2f7Ix9-dn9r9gmxD62CM7npGS5yB3bw3A-0ecNmZDfZtwa4kMn2eiy_dvIgznUcgrnBhQGN7hP2QZHy$

    Though C14 seems strong for authenticity, yet, maybe minor point, it lacks margins (like “Jesus Wife”), and, perhaps much more importantly, why would Kando or Sa’ad gift or sell it to an (unnamed) Montana woman, who presumably could not read it nor realize its significance and value? They, at minimum, could recognize the script.
    If it were genuine, it would have been only the second then-known such paleo-Hebrew ms, the first was from a Wadi Murabba’at cave, published in 1961.

  27. tapping the brakes

    Seems as if that should always show earnest. However, some people have learned to tap the brakes occasionally for show, while continuing to race to conclusions. I find deviousness much more instructive than honesty.

  28. David Eddyshaw says

    Several references to “Museum of the Bible Forged Scrolls” initially garden-pathed me into imagining that there might be an actual Museum of Forged Bible Scrolls. (I think there should be …)

  29. I read it as a reference to a museum housing those scrolls forged by the Bible….

  30. First word read in a Herculaneum scroll! The word is πορφυρας.

  31. Yeah, I thought about posting that but then I thought “One lousy word? I’ll wait until there’s at least a phrase!” But thanks for linking it; it’s definitely LH material.

  32. >at least a phrase

    It looks like they’re getting close. Here’s an image of a longer segment.

  33. Wow. OK, that’s impressive.

    *starts waiting for new Sappho or Archilochus*

  34. I imagine it’s a new text, whatever it is, or someone would have recognized it already.

  35. This doesn’t really contradict you, Y, since the term text can encompass different meanings, but they may find more complete versions of texts only known partially, in epitome or in quotation by later authors, in which case the initially revealed passage might not be recognized.

  36. Oh, sure. I was thinking of ‘text’ with the possible meaning ‘a part of a composition’. I’m sure that as soon as anyone deciphers an uncommon word, they rush to the concordance to see if its context exists elsewhere.

  37. Radhika Rajkumar at ZDNet:

    On Tuesday, researchers at the University of Pisa in Italy announced that they have successfully used AI to decipher a papyrus scroll found in Herculaneum […] By identifying and translating 1,000 words, or roughly 30%, of the scroll, the team discovered Greek philosopher Plato’s final resting place: a garden at the site of the Platonic Academy in Athens. The text also reveals that Plato was sold into slavery in either 404 or 399 BCE — not 387 BCE, as historians believed prior to Tuesday’s discovery.

  38. That’s wild. I wonder why 404 or 399. I hope the text is published soon.

  39. Trond Engen says

    And it’s only the beginning. But how do they verify the findings without destroying the scrolls? I guess it’s as simple as doing it again with different software, assuming that the learning algorithms (and maybe the input) are independent.

    And then it’ll be tested over and over again with better imaging and new iterations of the software.

  40. I wonder why 404 or 399.

    Time for wild guessing! Probably it’s not either-or, but between. Between the fall of Thirty tyrants and the death of Socrates.

Speak Your Mind