Sentiment strength detection for the social web: From YouTube arguments to Twitter praise

This talk described simple methods for detecting positive and negative sentiment strength in the informal language that is common in the social web. The Java program SentiStrength was described, demonstrated and evaluated for English language text. SentiStrength was also applied to large scale social web text from Twitter and YouTube to show how its results can be used. The talk explained how SentiStrength is language-neutral but can be adapted to different languages by changing the linguistic input files and some of the algorithm parameters.