Ronen and co-authors from MIT, Harvard University, Northeastern
University, and Aix-Marseille University tackled the problem by
describing three global language networks based on bilingual tweeters,
book translations, and multilingual Wikipedia edits. The book
translation network maps how many books are translated into other
languages. For example, the Hebrew book, translated from Hebrew into
English and German, would be represented in lines pointing from a node
of Hebrew to nodes of English and German. That network is based on 2.2
million translations of printed books published in more than 1000
languages. As in all of the networks, the thickness of the lines
represents the number of connections between nodes. For tweets, the
researchers used 550 million tweets by 17 million users in 73 languages.
In that network, if a user tweets in, say, Hindi as well as in English,
the two languages are connected. To build the Wikipedia network, the
researchers tracked edits in up to five languages done by editors,
carefully excluding bots.
In all three networks,
English has the most transmissions to and from other languages and is the most central hub, the team reports online today in the
Proceedings of the National Academy of Sciences.
But the maps also reveal “a halo of intermediate hubs,” according to
the paper, such as French, German, and Russian, which serve the same
function at a different scale.
In contrast, some languages with large populations of speakers, such
as Mandarin, Hindi, and Arabic, are relatively isolated in these
networks. This means that fewer communications in those languages reach
speakers of other languages. Meanwhile, a language like Dutch—spoken by
27 million people—can be a disproportionately large conduit, compared
with a language like Arabic, which has a whopping 530 million native and
second-language speakers. This is because the Dutch are very
multilingual and very online.
The network maps show what is already widely known: If you want to
get your ideas out, you can reach a lot of people through the English
language. But the maps also show how speakers in disparate languages
benefit from being indirectly linked through hub languages large and
small. On Twitter, for example, ideas in Filipino can theoretically move
to the Korean-speaking sphere through Malay, whereas the most likely
path for ideas to go from Turkish to Malayalam (spoken in India by 35
million people) is through English. These networks are revealed in
detail at the study’s
website.
The authors note that the users they studied, whom they consider
elite because—unlike most people in the world—they are literate and
online, do not represent all the speakers of a language. However, “the
elites of global languages have a disproportionate amount of power and
responsibility, because they are tacitly shaping the way in which
distant cultures see each other—even if this is not their goal,” Hidalgo
says. When conflict in Ukraine flared this past summer, most people in
the world learned about it through news stories originally written in
English and then translated to other languages. In this case, “any
implicit bias or angle taken by the English media will color the
information about the conflict that is available to many non-English
speakers,” Hidalgo says.
The networks potentially offer guidance to governments and other
language communities that want to change their international role. “If I
want my national language to be more prominent, then I should invest in
translating more documents, encouraging more people to tweet in their
national language,” Ronen says. “On the other side, if I want our ideas
to spread, we should pick a second language that’s very well connected.”
For non-English speakers, the choice of English as second or third
language is an obvious one. For English speakers, the analysis suggests
it would be more advantageous to choose Spanish over Chinese—at least if
they’re spreading their ideas through writing.
The problem of measuring the relative status of the world’s languages
"is a very tricky one, and often very hard to get good data about,”
says Mark Davis, the president and co-founder of the Unicode Consortium
in Mountain View, California, which does character encoding for the
world’s computers and mobile devices. “Their perspective on the problem
is interesting and useful.”
Cultural transmission happens in spoken language too, points out
William Rivers, the executive director of the nonprofit Joint National
Committee for Languages and the National Council for Languages and
International Studies in Garrett Park, Maryland. Data on interactions
in, say, the souks of Marrakech, where people speak Arabic, Hassaniya,
Moroccan Arabic, French, Tashelhit, and other languages, are impossible
to get but important in cultural transmission, he says. He adds that “as
the Internet has become more available to more people around the world,
they go online in their own languages.” When they do, now they know how
to connect to other languages and move their ideas, too.
