Salience uses a phrase-based approach for sentiment. We have the concept of “intensifiers” and “negators”, so that we can have a reasonably sized dictionary across our various languages that can then deal with things like “good”, “very good”, “not good”, etc – by using combinations of sentiment phrases and intensifiers/negators.
In 5.1.1 we added the concept of “superlative” scaling, so that, if you have a word like “good”, that has a sentiment score of 1, the superlative form (best) will be scaled by 1.25x (default) to give you a score of 1.25. We also added “emphatic” scaling, so, if the sentence has an “!” at the end of it, the sentiment for the entire sentence is scaled by 1.5x. So, if you see “Lexalytics is the best!”, that sentence will be scored at 1 (good) x 1.25 (superlative) x 1.5 (emphatic) = 1.725.
Those default scaling values are, in the tradition of Lexalytics Salience, fully customizable as part of our continuously expanding Salience Options.