The Rosetta Stone was one of the most important documents in the history of linguistics. Discovered around 1800, it allowed Ancient Egyptian to be deciphered. Let's say that the stone didn't exist, was destroyed like the Library of Alexandria, or hadn't been found yet. Knowing that Ancient Egyptian is an Afroasiatic language, could the hieroglyphs have been deciphered using the other Afroasiatic languages which were known (Chadic, Cushitic, Semitic, etc.) and some hard cryptanalysis?

  • 10
    Remember that a big part of our modern understanding of the Afro-Asiatic languages rests on our knowledge of Egyptian.
    – Draconis
    Commented Nov 3, 2019 at 22:07
  • 2
    Are you asking only about the Rosetta Stone, or if no other trilingual records had been discovered?
    – curiousdannii
    Commented Nov 4, 2019 at 3:31
  • 5
    Do you think that Linear A could have been deciphered without a bilingual? Oh, wait a moment ...
    – Colin Fine
    Commented Nov 4, 2019 at 10:49
  • 8
    Cryptanalisys would be useless for this task, since unknown languages aren't written in code; the goal of cryptanalisys is to recover an unknown plaintext given a known cyphertext (the specific algorithm and the encryption key are unkown). The book "Cracking Codes: The Rosetta Stone and Decipherment" (R. B. Parkinson, Whitfield Diffie, Mary Fischer, R. S. Simpson) explains clearly the difference between decipherment and cryptanalisys.
    – Vorbis
    Commented Nov 4, 2019 at 14:59
  • 4
    @Vorbis: I disagree (mildly) with your remark on crytanalysis: M. Ventris was a cryptanalist and used cryptoanalytic methods to crack Linear B. Cryptanalysis can do sometimes something good when the writing system itself is unknown. It does not help when you have no idea of the language encoded. Commented Nov 5, 2019 at 12:55

2 Answers 2


Fundamentally, Hieroglyphic Egyptian was cracked by Champollion, using two hypotheses.

  1. Egyptian is the ancestor of Coptic.
  2. The names in the "cartouches" are royal names. These hypotheses gradually led to a consistent set of readings for the hieroglyphs, that ultimately enabled to read all inscriptions.

The fact that Egyptian belongs to Afrasian is a consequence of the deciphering, and played no role in the deciphering itself. In all cases, the branches of Afrasian are too different from each other to be of any help.

If you take Old Nubian, it's still very hard to read and understand, even though it's related to present-day Nubian. So difficulties remain, even when the language to decipher is a direct ancestor.

The advantage of the Rosetta Stone is that the same text is written in three different languages. Without such a document, deciphering would certainly be possible, but considerably more difficult and, what is more, hard to prove it's a real deciphering.

  • 4
    Does this mean that even with the stone, it would mostly remain undeciphered today if it had died out with no known descendants (like Etruscan or Minoan)?
    – T.E.D.
    Commented Nov 4, 2019 at 14:22
  • 15
    @T.E.D. Unfortunately yes. Examples are Etruscan that we can read, but only partly understand, and ancient Meroitic that we can read but not understand at all (though there are some statistical analyses suggesting that it is another Afro-Asiatic language, but that's the best we have about it). On the other hand, Sumerian and Elamic died out without descendants, but we have a good grasp of those two languages because of ancient dictionaries to Akkadian. Commented Nov 4, 2019 at 15:46
  • 4
    @T.E.D. Scripts like LinearA and Indus script are indeed very hard to decipher because it is not clear which languages are written in these inscriptions, apart from the fact that inscriptions are very short. So if we did not have Coptic, hieroglyphic Egyptian would probably remain undeciphered.
    – user23769
    Commented Nov 4, 2019 at 18:12
  • 3
    @jknappen-ReinstateMonica what does it mean to read but not understand? Commented Nov 6, 2019 at 9:25
  • 9
    Ancient Meroitic is written in an alphabetical script. So one can literally read it aloud without understanding. Same for Etruscan. Commented Nov 6, 2019 at 10:50

Possibly (but probably not)!

Translating between two languages with only monolingual corpora (unsupervised machine translation) is currently possible. It's an area of active research in NLP because current machine translation methods use large, parallel sentences which are expensive to create and don't exist between many language pairs.

The current state-of-the-art in unsupervised translation is Song et al. (2019), which reports BLEU of 37.5 on English-French. For reference, Google Translate, which uses parallel data, only scored about 35.7 as of 2017 (higher BLEU is better) (Johnson et al., 2017).

However, EN-FR is one of the easiest pairs because:

  • There is a lot of high-quality parallel and non-parallel data since both are official languages of the U.N., the E.U., various countries, etc., are spoken by millions of people worldwide
  • The languages have many cognates and some shared vocabulary
  • The languages share a fairly simple writing system (esp. compared to hieroglyphics)

Lample et al. (2018) tested their system on Urdu->English, two unrelated languages with different writing systems and with (relatively) little available data, and obtain 12.3 BLEU. I don't have a reference point for how good that is, but it's definitely a start.

Finally, Zhang et al. (2019) train a translation system on Chinese -> Japanese, and show that it is possible to learn information about logographic writing systems, but Japanese kanji is borrowed from Chinese characters, so there's a lot of shared vocabulary.

That being said, hieroglyphs are (IMO) a more complicated writing system than even Chinese. And even for English-Urdu, a "low-resource" language pair, Lample et al. use 5.5M sentences. I have no idea how much text exists in hieroglyphs, but suspect it's less than this. But in theory, if we dug up and digitized millions of tablets of an ancient, unknown language, then yes, we do have tools to translate it.

  • I think that the OP was technically trying to ask about machine decipherment rather than machine translation, ...? Commented Nov 5, 2019 at 13:16
  • 2
    @RBarryYoung Unsupervised machine translation is relevant to this: it can (in principle) produce the bilingual text we are missing otherwise. But note the requirements on corpus size for training! Commented Nov 5, 2019 at 13:26
  • Ah, thanks, good to know. Commented Nov 5, 2019 at 13:33
  • 2
    Picking nits here, but Japanese and Chinese share very, very few cognates if any at all. What they do have is an enormous set of vocabulary borrowed from Chinese into Japanese and a smaller set borrowed in the opposite direction, but those are loans, not cognates. You could argue that modern loans from other languages (like Ch. 巧克力 and Ja. チョコレート, both from English chocolate) constitute cognates, but it’s stretching it a bit and probably wouldn’t even be useful for this. Commented Nov 5, 2019 at 19:36
  • 1
    @AlexeyRomanov - From my experience with Wikipedias in other languages, Urdu Wikipedia is likely to have a lot of text produced by Google translating (and hopefully editing) articles from English Wikipedia and other languages. Including machine translated text in a corpus intended to test machine translation doesn't sound like a good idea.
    – Pere
    Commented Nov 6, 2019 at 21:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.