The method uses a list of words and their average sentiment orientations, together with a set of additional rules that make use of the different ways in which sentiment is expressed. For instance, one rule increases the strength of any detected sentiment if a word is deliberately spelled with additional letters (e.g., haaaaaappy) and another rule reverses the polarity of any detected sentiment in the presence of negating words (e.g., not). The talk will give illustrate a large-scale social application with an analysis of important media events via Twitter posts.