Found in Translation

Me dical Inte rpre te rs Save Live s—and Mone y
A 2010 study showed that professional interpreters improved efficiency and throughput in
emergency rooms while decreasing the patient’s overall length of stay. The reason? When language
barriers exist and no interpreters are available, healthcare providers are more likely to order
expensive diagnostic tests to determine what is wrong with patients and monitor their care for longer
periods than necessary, resulting in excess spending. The same study also found that patients who
had professional interpreters were four times more likely to be satisfied than patients who did not.
5
Out-Translating the Outbreaks
You’ve probably seen this movie: Doctors with gas masks and hazmat suits
staring into microscopes at rapidly reproducing microbes. Somber politicians
viewing graphics that illustrate how quickly the rogue virus will spread if not
contained. The desperate search for an antibody. It’s the ultimate Hollywood
fearmongering scenario of mutant microbes taking over the world. And then
suddenly, like during the height of the SARS outbreak in 2003 or the swine flu
outbreak in 2009, parts of the horror movie script become reality on the
nightly news. When such things happen, we use hand sanitizer, purchase face
masks, cancel our flights, and huddle close while the threat lasts. But in
between these alerts, most of us tend to forget the lurking menaces. After all,
there’s not much we can do anyway, right?
Well, fortunately not everyone despairs. Behind the scenes, scientists work
diligently to prevent epidemics, enabling the public to regain peace of mind.
How does translation play a part? In 1997, after the first outbreak of bird flu in
Hong Kong, two physicians from the Canadian federal health department had a
vision for prevention. They knew that local media sources often reported
homegrown epidemics before the national or international media became
aware of them. So, they developed a system to automatically monitor that
media in real time, look for certain keywords, and send alerts based on the
occurrence of those important terms.
6
This early alert system, called the Global Public Health Intelligence Network
(GPHIN), leads to quick detection and response that can help prevent a wider
outbreak. The Canadian physicians developed GPHIN in partnership with the
World Health Organization (WHO), processing worldwide news sources in
French and English to detect local news reports that mention human diseases as

---

well as animal and plant diseases; biological, chemical, and radioactive
incidents; and even natural disasters.
There was just one glaring problem with that early concept: They were
looking at media sources in only two languages, English and French. Clearly
there was a need to include more languages. After all, chances were slim that
the local high school English teacher in the Chinese hinterland would take it
upon herself to file a report in English on a disease outbreak in her township.
So, the GPHIN developers quickly added Arabic, simplified Chinese (the
written form of Chinese used in mainland China) and traditional Chinese (the
written form used by most of the rest of the Chinese-speaking world), Farsi
(Persian), Portuguese, Russian, and Spanish.
Since its official 2004 launch at the United Nations, the GPHIN system has
been retrieving news items every fifteen minutes in those nine languages from
news aggregators Al Bawaba and Factiva. It translates the articles from those
languages into English, and also translates the English articles into the other
eight languages. GPHIN’s developers soon realized that the daily diet of
approximately four thousand original articles with potentially relevant content
could be handled only with a mixture of computerized or machine translation
and appropriate human oversight. So the developers chose several software
programs to automatically translate information in the various language
combinations. Once the articles are machine-translated, the system either
rejects them as irrelevant, flags them for analysis by humans, or publishes
them immediately to alert the worldwide meta-government and government
subscribers of a potential threat.
The Canadian system is still working to keep your neighborhood outbreak-
free today. As examples of the system’s vigilance, both the swine flu and the
SARS epidemics were first discovered by GPHIN. In both cases, the alerts
launched a process of response and containment that significantly decreased
the severity of the outbreaks.
At first, this may sound like a perfect out-of-the-box solution. Once set up,
the program operates indefinitely without too much additional work, right?
Well, not exactly, say the developers. Algorithms need ongoing fine-tuning to
match the constantly changing developments in each language. For instance,
the Chinese term for AIDS is
(aizi bing). Its first part, aizi, is a
transliteration for “AIDS,” and the second part is a classifier for disease.
Locals often use unofficial terms, such as
,
, or
, which are
all pronounced identically but also mean “the disease caused by love,” “the
disease of loving capitalism,” or “the disease of loving oneself.” The GPHIN
system must pick up these unofficial terms and translate them properly too, and

---

that’s only possible if it is continuously trained by human specialists.
As another example, take the article titled “Yellow Fever” that was published
in the Tampa Tribune in 2003. Here’s an excerpt: “An epidemic of penalties has
thwarted many drives, resulting in a three-game losing streak and essentially
leaving the Bucs’ season on life support.”
7
Ring the alarm bells: There are
plenty of terms here to cause the GPHIN system to go into overdrive. However,
if the algorithms are smart enough to offset those words with the sports-related
terms in the same sentence, the system will not automatically sound an alert.
Instead, it hands the report over to a human analyst who can quickly put the
information into the right context.
So when you find yourself rolling your eyes at the bizarre automatic
translation renderings you find on the web, remember that there are some uses
of computerized translation tools that are saving lives. And in the process,
translation keeps the hazmat suits and political disaster scenes on the screens of
Hollywood and out of our daily lives.