Microsoft Suggests It Fixed Chinese Bing 'Bug'
|
On Nov. 20, Microsoft took a broadside from New York Times columnist Nicholas Kristof over Bing. The two-time Pulitzer Prize winner suggested in a column that Bing delivered uniformly pro-Chinese-government results to politically sensitive queries inputted into Simplified Chinese. I wrote about that original controversy here. Kristof said in his column that Simplified Chinese results in Bing were scrubbed "wherever in the world the search is conducted--including my office in New York." Meanwhile, he suggested, the use of "complex Chinese characters (the kind used in Taiwan and Hong Kong)" gave more impartial results. Microsoft, of course, felt duty-bound to respond, suggesting in a Nov. 20 posting on the official Bing blog that "some queries ... provide very balanced Web results" and that "we can continue to improve our relevancy and comprehensiveness in these results and we will." However, "today's investigations uncovered the fact that our image search is not functioning properly for queries entered using Simplified Chinese characters outside of the PRC," Adam Sohn, senior director of Bing, wrote in the blog posting. "We have identified the bug and are at work on the fix. We expect to have this done before the Thanksgiving holiday." On Nov. 30, I decided to see for myself whether the bug was fixed, using a combination of friends fluent in Chinese, along with Google Translate as a fact-checking backstop. Below, I'm giving a run-through of my process. When you input the term "Tiananmen Square" into Bing's Image Search in Simplified Chinese characters, your results page looks like this:
If you scroll down, you'll see they're all postcard-style shots, most of them from Xinhua, the official press agency of the PRC. Enter "Tiananmen Square" in English, and you get this:
As you can see, these are mostly images from the 1989 protests, including that much-reprinted shot of a lone citizen standing in front of a line of tanks. In that official Bing blog, Adam Sohn asserted that "we did fix a bug in web search" and pointed to the Simplified Chinese search term "June 4th Tiananmen" (the date of the 1989 protests) as an example of a term that gave new, balanced results. Indeed, as the image below shows, inputting that term results in a variety of exceptionally brutal photos from the protest, including that iconic man-vs.-tanks:
I wanted to input a similar term for the protest/massacre, choosing "Tiananmen Incident." I received this:
There are a couple of protest-related shots here, but they seem more mixed in with random historical images. Next, I wanted to check the results if I typed a politically sensitive term into Web Results. I started with "Falun Gong," and got this:
Note the mix of pro-Chinese-government sites (Chinamil.com.cn; people.com.cn) along with pages that present a somewhat more balanced view, such as zh.wikipedia.org. After that, I inputted "Tiananmen Incident" in Simplified Chinese:
Some of these sites are pro-Chinese-government, offering a more sanitized perspective, but others--such as boxun.com--contain a more balanced view, along with images of man-vs.-tanks:
A Microsoft spokesperson told me in an e-mail on Nov. 30 that "the bug identified in the web image search was indeed fixed," and then went on to reiterate a quote from Sohn: "Please also note that Microsoft 'recognize[s] that we can continue to improve our relevancy and comprehensiveness in these web results and we will.'" With the exception of my original "Tiananmen Square" image search, Bing does indeed seem to offer relatively balanced results pages for politically sensitive queries in Simplified Chinese (at least in my opinion). What do you think? Do you agree with Microsoft that the bug has been fixed? |









Comments (3)
I read somewhere that Bing weight pages that are written in the same language as the search query higher than other results. I bet that has a fairly large effect on this also.
Posted by Erik Jorgensen | December 3, 2009 2:28 PM
I don't see how this is anything more than a complete misunderstanding by the NYT of how web search works, or how It should be expected to work.
In English the term 'tianenmen square' on a web page is most likely to refer to the massacre, so it makes sense that the search results reflect that.
However, in Chinese, the translated 'tianenmen square' term appearing on a web page is very unlikely to refer to the massacre, since for most of the web population using that language, posting as such is likely to get you shot in the back of the head.
So, given the relative weightings of pages using the two terms in the two languages, I'd say the results make perfect sense.
There are probably a whole set of example of this idiomatic variance. But since this one involves politics correctness it has to be an important story, right? Anything to sell dead trees...
Posted by Piersh | December 11, 2009 11:52 AM
Bing is a decision making engine. What kind of decision you plan to make when you search a place like 'tiananmen square'?
If I were a visitor, I would like to see results about its present condition, surrounding, traffics and attractions. History of the place would be second in my decision making. Other people who use Bing to search this word seems to be interested not only in history, but also in maps, posters and gates as well. Maybe people who use Bing and Google represent different populations, I think it has something to do with Google's lack of stats from Chinese mainlanders who are unable to access Google
Posted by jason | January 13, 2010 7:56 PM