The number of user reviews available online is constantly growing, automated extraction and summarisation of the useful information is of critical importance for both businesses and customers. For example, suppose you have hundreds of reviews on a specific model of a digital camera. Typically, such reviews would contain both positive and negative feedback on various features of the camera: for instance, the majority of the users might agree that picture quality is good, but battery life is poor. On the one hand, if you were a data analyst working for the camera producer, you would conclude that it is improvements in the battery life that the company should focus on primarily; on the other hand, if you were yourself a customer thinking of buying a new camera, the information about different positive and negative aspects of the product might influence your decision.

Using reviews extracted from Amazon and manually labelled with product features and sentiment polarity as well as sentiment strength, I propose a system to automate extraction and summarisation of the useful information in reviews.