Robert Munro (@WWRob) is a graduate fellow at Stanford University and chief executive at Ibidon who’s fascinated with languages that are spoken by very few. At Where 2012, Munro will examine differences in how people express place, distance, and space among the world’s 5,000-plus languages. These are differences that location-aware app designers will have to pay attention to as the world’s data becomes less and less predominantly English.
Our interview follows.
What are some of the ways that space and direction are expressed differently among cultures?
Robert Munro: We tend to think of spatial distinctions as absolute and clean-cut, but when you scratch the surface, it is clear that we encode space and direction in many very subtle and language-specific ways. If I said, “look in front of the house,” do I mean “in front of” relative to you, me, or the house itself? Or if we’ve been looking at an aerial map together, maybe I just mean to the north, south, or roadside.
Even in English, location can be difficult to embed into technology. If you ride the light rail in Portland, Ore., you might hear an ethereal voice saying, “The doors on my right side will open.” Is this the (somewhat creepy) disembodied voice from the train that you are inside, meaning the doors on the right in the direction of travel? Is it from the virtual driver facing the passengers, meaning the doors on the left in the direction of travel? Portland recently hosted the annual conference of America’s top linguists; even they had to wait for the train to stop to figure it out.
Why are these differences important for geo developers to consider?
Robert Munro: The majority of the world’s digital information is now in non-English unstructured text. There are some 5,000 languages in the connected world, each of them spoken by people operating potentially location-aware devices.
The location-based applications on those devices are abstracted from space, time and
direction, using visual and verbal metaphors for each. Every language
encodes space and time differently and there is a direct relationship
between visual spatial metaphors and language. Recent research has
found this link to be much stronger than once thought — even short-term changes in language can drastically alter the way we perceive
space and time. For some languages, the future is in front of us, but
for others it is behind us. Some languages use length is a metaphor
for time (a long time) while some prefer volume (a large amount of
time). Some languages will prefer relative directions (turn left)
while others will only permit absolute directions (turn north).
Location-based applications use every one of these metaphors, so it is
important for developers to understand the rich breadth of spatial and
temporal metaphors that they are (often unconsciously) coding into
their applications. As the applications often have a visual language
of their own, it is safe to say that they can also have a direct
influence on their users’ sense of space and time. After staring at a
map before exploring some new place, did you ever feel like you then
had an augmented reality view of the place thanks to that map? In a
sense, you probably did.
Can you tell us about the work you did after the Haiti earthquake to crowdsource local information from a global diaspora of Haitians?
Robert Munro: Mission 4636 was a crowdsourced information service established in the wake of the 2010 earthquake in Haiti. People within the country were able to send free text messages that were then translated and mapped by crowdsourced workers and streamed back to responders within the country. We launched in 48 hours. I ran the actual crowdsourcing component — finding and managing thousands of Haitian Kreyol and French speakers globally who could help.
It taught me two important lessons. The first is that nothing beats local knowledge. Members of the Haitian diaspora thousands of miles away were quicker at geolocating vital locations in Haiti than many international workers on the ground. The second lesson that it taught me is that it is hard to find those who have the crucial local knowledge. At the time, I needed to reach out through social media, looking for (mostly) ad-hoc groups of people with Haitian affiliations. There simply wasn’t a unified resource that links people by languages spoken.
This has influenced me greatly. For every corner of the world, there is a global network of largely untapped local knowledge, and it is just as applicable to travel and trade as it is to crisis response. It gives us one solution to the problem of encoding location in technology. For a given problem, there is probably someone online right now who does have the necessary local linguistic and geographic knowledge. In addition to making technology smarter across languages, we also need to be making technology that can better engage these individuals. With Idibon, we are doing both, building language-processing algorithms that can parse unstructured data in previously unknown languages and linking this to a global network of language speakers. In Haiti, we were able to transfer the system to local, paid workers within the country, continuing the local knowledge and providing employment where it was needed most. I would love to see this kind of approach adapted more broadly, turning people’s knowledge of less well-known languages to their advantage.
How is location used in viral forecasting by health organizations?
Robert Munro: Location is the oldest and newest component in viral forecasting. In
the 1850s, John Snow founded both modern epidemiology and geographic information systems (GIS) when he mapped the location of a London cholera outbreak to a single water pump.
Today, unfortunately, the dynamic maps that you see in outbreak movies outpace the technologies of actual epidemic-tracking organizations. But this is rapidly changing on the back of communication technologies. While HIV took decades to be isolated, Bird Flu and Swine Flu were isolated weeks or months after they first appeared. We can go back and find early, open reports about Bird Flu and Swine Flu with the key signatures of a new virus. This could have enabled us to identify and contain them even earlier, and smarter information processing will allow this for future outbreaks. Within Global Viral Forecasting, a spin-in called epidemicIQ is doing just this, looking to track outbreaks as early as possible. (Disclosure: I was the CTO of epidemicIQ until recently and continue as an adviser.) Local knowledge is vital here, too: 90% of the world’s outbreaks come from 10% of the land; 90% of the world’s languages also come from 10% of the land. It is the same 10%.
Geographic information systems are finding their way back into real-time outbreak tracking, too. Open data about flight paths and transportation networks allows us to track the potential spread of an outbreak with a precision that was never before possible. At the moment, we are quickly closing the gap on how quickly we can track the spread of outbreaks. Will we soon be able to use the same technology to get ahead of outbreaks? I hope so.
Who else is doing interesting work in this area?
Robert Munro: For almost half a century, most language researchers assumed that the cross-language differences in space and time were more or less arbitrary. Within the last few years, a number of researchers, the Stanford psychologist Lera Boroditsky in particular, have produced strong evidence showing that they can manipulate people’s sense of time and space by changing their language, even over very short periods. It has turned many people’s assumptions about language and world-view right around.
In health, Nicholas A. Christakis at Harvard and James Fowler at the
University of California San Diego have made some very interesting discoveries about the spread of disease, using online social networks as a powerful stand-in for geographic/interpersonal connections. They have been able to make reliable predictions about the spread of outbreaks and other communicable phenomena to several people out along social networks. It is a really exciting new approach.
This interview was edited and condensed. Photo: I’m So Confused! by Ian Sane, on Flickr