Sentiment and Content Analysis to cluster neutral messages online

The purpose of this research is to apply both sentiment and content analysis methods to neutral messages posted online. Past studies have revealed that the classical method adopted to conduct sentiment analysis has important limitations. First, neutral messages are often considered "good-for-nothing" material or literally something that tools are not yet able to classify. However, some new studies have shown the importance of considering neutral messages as a proper category with its own aspects because of its potential for improving the accuracy of positive and negative classifications. This paper aims to articulate a more reliable method for understanding neutral posts, based on a combination of sentiment and content analysis; then provide new "labels" for the creation of ad-hoc clusters of neutral messages. By doing so, we contribute to the discussion in online content analysis depth and analysis methods and represents one piece of a larger research project examining the quality of e-relationships as expressed through online content.