Will the Indus Valley script ever be deciphered?

Around 4,000 years ago, one of the world's oldest civilizations emerged: The Indus Valley Civilization, flourishing in what is now Pakistan, western India, eastern Iran and parts of Afghanistan. In addition to building sizable cities, its people created a written script that consists of hundreds of signs that remain undeciphered.

The signs, sometimes called Harappan script, vary, with some looking like a diamond with a square in its corner; a U with three "fingers" at each end, and an oval with an asterisk-like shape inside it.

Most of the surviving texts are inscribed on nonperishable materials, like clay and stone. So will this undeciphered script ever be deciphered? Could recent advances in artificial intelligence (AI) help with decipherment?

The undeciphered script

The Indus Valley Civilization flourished between roughly 2600 and 1900 B.C. Thousands of artifacts containing the script survive to the present day, Michael Philip Oakes, a researcher in computational linguistics at the University of Wolverhampton in the U.K, wrote in a paper published in the Journal of Quantitative Linguistics.

The surviving texts tend to be very short, with an average of five signs per text, Oakes noted. There is no known bilingual text recorded in the Indus Valley script and a known text to aid with decipherment — in other words, Indus Valley Script doesn't have its own Rosetta Stone. It's also uncertain which language the script encodes, and some scholars have argued that it may not encode a language at all, suggesting that they may function more like emblems that convey a person or entity.

Exactly how many signs the script contains is a matter of debate, but they number in the hundreds, Oakes said.

Experts have a mix of ideas about whether the script will ever be deciphered. Even if it is decoded, the texts' short lengths and scholars' wide differences of opinions may make it hard for any decipherment to be widely recognized.

While some experts think that AI could help decipher the language, researchers will likely have to guide the AI for a full decoding, the experts said.

Is it partially deciphered already?

Steve Bonta, an independent researcher who holds a doctorate in linguistics and has studied the script extensively, said part of the work may already be done.

"I think that the Indus Valley Script is already partially deciphered, but that recognition of that fact is severely lagging," Bonta told Live Science in an email. Bonta said he showed "back in the '90s that certain signs and canonical sign fields must be indicative of notations of assets, expressed in various weights." However, many scholars do not recognize the decipherments as accurate.

Bonta said his claims of partially deciphering the script are far from alone. Prior to the mid-90s, "claims of decipherment were published fairly regularly," Bonta said. None of these claims has gained widespread acceptance, with one problem being that the shortness of the surviving texts makes it difficult to prove the accuracy of any decipherment.

"Most of the Indus inscriptions are brief and highly repetitive, which makes the task of reproducible decipherment very difficult," Bonta said.

Turning to AI

AI is useful for decipherment attempts and may help researchers generate lists of possible sign values. However, in the end, human researchers will still need to take the lead. AI "is an extension of human intellect and intuition, albeit an extraordinarily powerful one," Bonta said.

Peter Revesz, a professor of computing at the University of Nebraska-Lincoln who is an expert in computational linguistics and has studied the Indus Valley script extensively, believes the script will be deciphered and that AI may play a significant role. Revesz's team has used data mining and statistical analysis to help determine which Indus Valley script signs are likely to have similar meanings.

The "Indus Valley Script surely will be solved one way or another, and AI can help, but it needs to be guided by a good research design," Revesz said in an email.

Rajesh Rao, a computer science professor at the University of Washington in Seattle who has co-written several papers on the Indus script, is less optimistic about a full decipherment but said AI will be useful. Back in the 2000s, with the more primitive AI that was available at that time, his team determined that the script has a statistical pattern that suggests that it encodes a language.

However, even with AI, a full decipherment seems unlikely with the existing texts, according to Rao. The chances "are not very high," Rao told Live Science, noting that a partial decipherment may be possible. "We may be able to reconstruct their number system," Rao said.

Rao said that the number system is already partly understood because some inscriptions have tally marks (vertical lines) that are thought to represent numbers. These are located next to symbols that likely represent objects. Additionally, archaeological data indicates that people used a system of standardized weights that involved ratios of 1, 2, 4, 8, 16, 32 and 64. By using the tally marks and weight system, it may be possible to determine which numbers are recorded on the inscriptions.

To decipher the entire script, Rao thinks archaeologists will need to uncover more texts. There are many Indus Valley Civilization sites that are largely unexcavated, and he hopes future excavations may yield lengthier texts or ones that feature the Indus Valley script alongside a known language.