Analyzing Sina Weibo (Chinese Twitter) and Renren (Chinese Facebook) for sentiment are quite tricky. In general social media analysis, for instance for Russian is tricky. There are few reasons:
- Grammar: in short messages there is not much space to spell out correct grammar. So in most cases it is “broken” from the stand point of classic parsers.
- Words: they change frequently, following social media development of a particular news / reaction or may be even a flash mob.
- Sarcasm: the author does not mean the sentiment you deduce by reading it for the first time. It sometimes takes a research and find a visual item, that helps understand the sentiment:
Fuxi API is catching up with what’s cooking in Chinese social media by analyzing a vast array of messages in Simplified and Traditional Chinese. We have just released its 1.2 version with a number of changes to better tune for the sentiment signal in the avalanche of tweets, blog posts and news articles, all in Chinese. Check it out.