How a professor created a new science vocabulary for an Indian language

Four years ago, when Siddhartha Jayanti began writing the first computer science paper ever published in Telugu, an Indian language with around 100 million speakers, he realized he had to first solve a problem that had little to do with computing—he had to find the words.

His research, published in English on arXiv in 2022, examined how groups of computer processors work together by looking at what each processor knows and how they communicate. Jayanti, now an assistant professor of computer science at Dartmouth, showed that there are fundamental mathematical limits on how quickly multiple processor cores can safely coordinate and share common data, by analyzing how information flows between them.

But Telugu, Jayanti’s native language, had no established vocabulary for the technical terminology he needed to communicate his results, such as “shared-memory multiprocessor.”

Now, in a recent paper on arXiv, Jayanti lays out how he turned to Sanskrit, the classical Indian language with an extensive influence on Telugu, to invent a new technical vocabulary that he sees as a first step toward building a scientific lexicon for advanced computer science scholarship across all Indian languages.

The original paper, initially a Telugu-language chapter in Jayanti's PhD dissertation, later won the 2023 ACM Principles of Distributed Computing Doctoral Dissertation Award as the first research paper in modern computer science written in an Indian language.

“When you can derive terms natively, they mean things to people, they can relate them to other words, they can understand them in a linguistic way, along with in a scientific way,” says Jayanti, who learned both Telugu and Sanskrit as a boy and was particularly interested in reading and writing poetry in both languages.

With more than 1.47 billion people, India is the most populous country in the world and home to dozens of languages, with Telugu being one of 22 official languages recognized by the government.

It's an enduring practice across Indian languages to take words from classical languages such as Sanskrit and make them their own through minor alterations of word endings or sound substitutions, Jayanti says: “This means that you can create a common vocabulary across a slew of Indian languages all at once.”

In his recent paper, Jayanti explains how he derived new words through simple examples such as “asynchronous,” which is used in the field of distributed computing to indicate that different computers or processors do their work independently without waiting for the other systems to stay in sync.

He adapted the Sanskrit term asamakālika from a- (not), sama (same), and kāla (time) to asamakālikamu in Telugu, in keeping with the grammatical rules of the language.

“Incidentally, the English term ‘asynchronous’ also comes from combining the Greek morphemes a- (not), syn- (together), khronos (time),” Jayanti notes in his paper.

The most complex term he reports, "shared-memory multiprocessor," became saṁvibhakta-smr̥ti bahusaṁsādhakamu, after Jayanti modified root words from Sanskrit with prefixes and suffixes to arrive at the translation.

Another hurdle in writing a computer science research paper entirely in Telugu was the inadequacy of existing software tools. Jayanti found that many programs did not fully support the language, with problems such as missing keyboard characters, fonts that displayed Telugu incorrectly, and typesetting software that could not properly handle the script.

He experimented with different tools, customizing and combining to create a complete workflow he called TeluguTeX, a system that allowed Jayanti to write both Telugu text and mathematical notation correctly.

Improving fonts, keyboards, and typesetting tools will make it easier for researchers to publish technical work in their native languages, he says.

Besides being the first modern computer science paper in Telugu, the 2022 paper, which established fundamental mathematical limits on safe multicore coordination, has also informed further research in the field of distributed and parallel computing. Working with collaborators at Princeton and MIT, Jayanti leveraged his results to develop highly efficient algorithms for a widely used data structure—a method computers use to organize and store data for fast access and updates. The algorithms can perform many operations at the same time, almost as fast as is theoretically possible.

Google now uses these algorithms in its open-source software for analyzing massive networks, he says, enabling it to process graphs with tens of billions of connections.

A scientific vocabulary for a digital age

Jayanti, who describes himself as "passionate about Telugu and Sanskrit," sees the Telugu-language paper as a proof of concept. Technical vocabularies accessible to Indians in their own languages can enable precise scientific expression, expand the scope for publishing scientific texts in regional languages, and provide a unified parlance while respecting the country's vast linguistic diversity and vernaculars, he says.

“Technology and science have become such a central part of our life. There should be a natural and easy way of taking on science, and what better way of trying to do that than to find a mechanism for making the research frontier of science and technology accessible in all of these languages,” Jayanti asks.

Jayanti proposed the Samskrtam Technical Lexicon Project as part of his PhD thesis and has since been involved in efforts related to the 2020 Indian National Education Policy. “Textbooks are being written at the central level in all subjects, both in regional languages and in English, and I am part of the group curating 11th and 12th grade computer science textbooks,” he says.

Eventually, the effort to create a scientific and technical lexicon that underpins more widespread science communication and research writing in Indian languages must be a collaboration between scientists and language experts, Jayanti says. The overlap between these expert communities remains small, he says, “but I think if we do this work, it can be very impactful to a lot of people.”