Mining User-Generated Financial Content to Predict Stock Price Movements

Sentiment extraction from user-generated online content to predict stock price movements has become an active research field. This paper gives an overview of common approaches to this topic and analyzes the content generated by the financial social network Seekingalpha.com. The first finding is that a large proportion of users’ attention is focused on only a few stocks. Regarding these stocks it can be shown that sentiment is significantly driven by past abnormal performance. Only the sentiment of premium users contains some degree of predictive
power. Generally, the users’ sentiment is consistent with a naïve
momentum mentality.