Data Visualized

Weibo, China’s Twitter, Abuzz with Sentiment Over Liu Xiang’s Olympic Fail

Posted on August 10, 2012 by SocialFlow

Liu Xiang’s 110m hurdles race was one of the most anticipated Olympic races for audiences across Mainland China. Shanghai born Liu Xiang has emerged as one of China’s most visible cultural icons, being the first Chinese athlete to achieve the “triple crown” of athletics: World Record Holder, World Champion and Olympic Champion. During the 2008 Beijing olympics Liu had to withdraw from the competition at the last moment due to a previously unrevealed injury. In this past week’s hurdle race Liu pulled his Achilles tendon while taking off. He attempted to jump over the first hurdle but crashed straight into it. He then hopped the full 110 meter stretch, helped by other athlete friends.

This drew numerous reactions online, and was heavily covered across western media outlets. However, only a few covered sentiment and reactions coming from Mainland China. This article posted on China Hush highlights some of the positive support coming from Chinese users, yet we wanted to see a more complete view of the story.

In the following post, we analyzed over 150k Weibo reactions to Liu Xiang’s race. We identified dominant words and user sentiment posted in reaction to the Chinese Olympian’s failure to complete the race. We see an uproar of support, but at the same time a wide range of critiques coming out of Weibo users in China. For the first time we have an ability to quantify and analyze sentiment from such a large population within China; a peek into the sentiment of China’s public sphere.

Sina Weibo

Sina Weibo has emerged as China’s most popular micro-blogging service, reported to have over 300 million registered users and about 100 million posts per day. It gives the public the ability to voice suggestions, reactions and critiques. The speed at which information spreads on the network makes it significantly harder to control. In many ways, Weibo is driving the entire national dialogue in China.

Weibo is similar to Twitter in many ways. Posts are limited by 140 characters and a similar one-directional friendship graph is used, hence users don’t need to reciprocate friendship when followed. Users are able to @mention other users on the network, as well as comment inline, repost or favorite someone else’s post. Hashtags aren’t as common (appear in 6% of posts) and when used, they appear in the following format – #Hashtag#. The service has some unique functionality, such as easily embeddable visual emoticons which can be attached to any post. Each emoticon has an explicit sentiment attached to it (such as ‘happy’, ‘frustrated’, ‘tired’, and so on) and once embedded in a post, spells out the chosen emotion in characters within the text of the post.

Over the past months we’ve been sampling Weibo’s public stream every few seconds using their API. While we don’t know the precise amount of public posts on the service at any given point in time, this method gives us a fascinating view into Mainland Chinese user sentiment, reactions towards specific events and how they evolve over time.

Reactions to Liu Xiang

We observed a total of 150k posts that contained Liu Xiang’s name (刘翔) between the early morning of August 7th (Beijing local time) and August 10th (early morning). The graph below represents the distribution of posts during the observed period, where the spike highlights the time when the race took place.

From these reactions, 11% included urls, 5.5% contained @mentions and 20.5% included one of the embedded emotions. The top most shared links were pages (such as this one) where users were able to weigh in on how they felt. The athlete’s bio page on Sina was heavily linked to as well as other pages with videos and images. Needless to say, the top shared urls were all services in China.

The following graph represents the most prominent nouns and adjectives shared in posts about Liu Xiang on Weibo. The larger the node, the more times it was used and the closer words appear together, the more frequently they were used by people in the same Weibo post. Once we organize the graph, we can clearly identify the language used in support of the athlete (blue: bottom right) and his critics (red: top left).

The public backlash is not surprising. After Liu Xiang’s previous episode at the Beijing Olympic games, many people accused him of being too busy performing in ads rather than training. After he pulled his tendon, a number of Weibo users accused Liu of pretending to fail. Phrases such as “movie“, “advertisement” and “able to win an Oscar” dominated posts with these views, heavily critiquing the athlete’s poor performance.

On the other side of the graph, we see numerous users coming out in support of Liu, using phrases such as “strong will”, “glory”, “proud” and “Olympic Spirit“. The edges and nodes in this part of the graph are much thicker, reflecting the fact that a greater number of users responded in a positive, supportive manner.


Tracking Sentiment on Weibo

Next, we mapped out the explicitly chosen emotions that Weibo users specified as they posted their reactions. We used the sentiment classification scheme outlined in this KDD Moodlens paper, and plotted the normalized number of times these explicitly chosen emotions appeared over time. Each emotion is plotted with a different color: excitement (blue), surprise (red), sadness (orange), doubt (green) and people who felt moved (purple).

The race took place around 5:40pm Beijing time which is when we see the drastic shifts in the graph (left side). Before the race there were substantially more people posting updates with excited emoticons. It doesn’t seem like folks were very much surprised (or they didn’t use the surprised emoticon while posting responses). While we see excitement drop almost instantaneously in response to Liu’s injury, at the same time we see an immediate spike in people’s usage of emoticons referencing sad sentiment. There’s a slightly slower yet consistent growth in people who use the moved related emoticons.

But note the even slower growing green line, representing doubt, growing over time and becoming even more (comparatively) prominent than those folks who identified as being moved. As observed by the entity word graph, there were substantially less words criticizing the Liu Xiang in comparison to those supporting him. This is reflected in the emoticon/emotion graph above, which clearly displays excitement, sadness and moved as dominant sentiment in response to the event. Finally, note the spike in excitement the following evening at 7pm and the decline in other emoticon usage that happened at the same time. Got any ideas why this happened?

In conclusion

When we look at Jack Moore’s Buzzfeed dashboard, it is clear that Liu Xiang’s fall has substantially more views, as well as the highest measure of virality (or viral lift, Buzzfeed’s defined metric for quantifying how likely a piece of content is to spread within the network) compared to other posts this week. The top referring websites were Facebook, Twitter and Reddit, linking to numerous posts across major media outlets.

In a similar manner, Sina Weibo is driving huge amounts of  traffic to pages across the Chinese web. Its’ networked power is derived from the speed at which information spreads between the hundreds of million users across China. For the first time ever, we have this incredible view into public perception in China, giving us the ability to identify a more nuanced public conversation.

Needless to say, we’re very excited by Weibo. Are you?

(This analysis is part of a summer project from one of our amazing Data Interns, Eddie Xie. Find him on Twitter | Weibo