Artificial Intelligence Stack Exchange is a question and answer site for people interested in conceptual questions about life and challenges in a world where "cognitive" functions can be mimicked in purely digital environment. It only takes a minute to sign up.
Sign up to join this community
Can Machine Learning be applied to decipher the script of lost ancient languages (namely, languages that were being used many years ago, but currently are not used in human societies and have been forgotten, e.g. Avestan language)?
If yes, is there already any successful experiment to decipher the script of unknown ancient languages using Machine Learning?
I would guess no, because if the language is unknown (no data available on it), then we would not have training data with which the machine learning algorithm could learn from.
If it is related to some known language, then some statistical analysis can lead to a guess at decipherment (assuming certain similarities among the two languages).
If interested on general language decipherment, see the following where they decipher scripts using available information on the language of interest: http://www.aclweb.org/anthology/W99-0906. They utilize the Expectation Maximization algorithm.
I'm sure more google searching can lead to other examples that use machine learning algorithms, but they would most likely have some known information or a body of assumptions to make our problem scope easier.
I say yes it definitely could be. But i agree with Skim you need some information as a starting point. Egyptian hieroglyphs were only (recently) understood following the discovery of the Rosetta Stone (https://en.m.wikipedia.org/wiki/Rosetta_Stone). With the same message in both known and unknown language the program could find the/a correlation. Without that info you would have to guess the potential content of the message. The results would then be confirmation-biased: how could you know it had worked properly? Say the program managed to ‘translate’, outputting a coherent phrase about bananas, great. Even if it was very consistent across multiple samples and 99% confidence, without any “control set data” it could just as easily be about, say, fish or even a prophesy on the super-bowl winner. Still a fun project tho.
There is an academic paper here that studies a neural approach to deciphering ancient languages:--
(https://arxiv.org/pdf/1906.06718.pdf)
"In this paper we propose a novel neural approach for automatic decipherment of lost languages. To compensate for the lack of strong supervision signal, our model design is informed by patterns in language change doc-umented in historical linguistics. The model utilizes an expressive sequence-to-sequence model to capture character-level correspon-dences between cognates. To effectively train the model in an unsupervised manner, we innovate the training procedure by formalizing it as a minimum-cost flow problem. When applied to the decipherment of Ugaritic, we achieve a 5.5% absolute improvement overstate-of-the-art results. We also report the first automatic results in deciphering Linear B, a syllabic language related to ancient Greek, where our model correctly translates 67.3% of cognates." ― Luo, Jiaming, Yuan Cao, and Regina Barzilay. "Neural Decipherment via Minimum-Cost Flow: from Ugaritic to Linear B." arXiv preprint arXiv:1906.06718 (2019).
Further Reading and Articles for the Layperson:--
I would say that it depends on whether that language would have wanted to be decyphered.
The origin of Cryptography dates back to around 700-800 AD. We may not know the methods by which these texts are obscurred. One such example is the Lesser Banishing Ritual of the Pentagram which, revealed only by happenstance, formed the basis of a whole new esoteric renaissance in the west, the lexicon included forming a basis for the lexicon in use by these orders today.
This Rite centers around the letters INRI, which tradition says were written upon the cross of Jesus Christ as an abbreviation for Jesus of Nazareth, King of the Jews. There are, however, numerous other levels of occult meaning regarding these four letters in the Rosicrucian Magical Tradition. One of these is a Hermetic secret alluded to by the Latin phrase "Igne Natura Renovatur Integra" which means "By fire, nature is perfectly renewed." These four letters additionally adorn the rays of the angles of the Rose Cross Lamen worn by Adepts of the Ordo Rosae Rubeae et Aureae Crucis.
A deeper interpretation lies occulted behind the attributions of the Hebrew letters and the Magical Forces to the Paths on the Qabalistic Tree of Life. The Path attributed to the Hebrew letter y is attributed to the Zodiacal Sign Virgo as well, that of n to Scorpio, and r to the Sun. There exist further magical associations between the Sign Virgo with the Egyptian Goddess Isis, Scorpio with Apophis, and the Sun with Osiris. When the first letter is taken from the Names of each of these Gods, the name "IAO" is formed. Additionally, due to the Signs associated with Isis, Apophis, and Osiris, they form the letters "LVX."
Thus within the letters IRNI lie concealed the letters IAO and LVX, which may also be found upon the rays of the angles of the Rose Cross Lamen worn by Adepts of the R. R. et A. C. The name IAO was considered by the Gnostics to be the Supreme Name of God. Its letters further allude to Salt, Sulfur, and Mercury in Alchemy and to an even more recondite secret symbolized by the relationship between Isis, Apophis, and Osiris.
So I would say for machine learning to be able to decipher real languages, it would almost have to be intuitive, able to draw parallels and connect various points of reference together. It would have to determine what things mean, when a lot of the times, this is virtually impossible.
Your variables are infinite. You have no real method of finding out what they are. Add to that context, lexicons, deliberate obfuscation, culture and the passage of time and I think it would not be realistic to expect machine-learning to provide much help.
