Differential Influence of Blogs Across Different Stages of Decision Making: The Case of Venture Capitalists

In this work, we followed the sentiment analysis literature, and used supervised learning methods, which take manually classified data (corpus) as input and automatically extract features (combination of words and parts of speech of words) for sentiment analysis (Dave et al. 2003; Ghose and Ipeirotis 2011; Pang et al. 2002; Shanahan et al. 2006). These supervised methods do not rely on manually or semi-manually constructed discriminant-word lexicons. Prior research has shown that supervised methods perform better than lexicon-based approaches for sentiment analysis (Chaovalit and Zhou 2005; Pang et al. 2002).