ViewPoint™ is configurable to measure sentiment at multiple granularities and/or dimensions. For example a product review posted on a news site can be analyzed holistically, at the paragraph level, or at the sentence level. In addition ViewPoint™ can score the sentiment from the point of view of a specific speaker or for a specific topic. Finally the sentiment scale can be defined while training a tailored sentiment analysis engine.

Out of the box ViewPoint™ has a default sentiment model that is trained on product reviews collected from the internet. Although this is an effective start if the domain of use is unknown or still being defined, it is better to train ViewPoint™ for the environment it is operating in. ViewPoint™ allows the user to specify a collection of documents labeled with a sentiment tag (i.e., positive, negative, or neutral) from which they can build a tailored sentiment engine.

If you are interested in the output of ViewPoint™, you will also want to look at the Document Classification and Clustering product to help you identify and/or discover topics in text to complement the sentiment scores.