Mining opinions on the basis of their affectivity

Recent reports reveal that users daily consult Web communities opinions for getting an idea about a product/service before they buy it. Furthermore, the same users, freely and without a direct economic return, collaborate in posting new reviews to the community. In most of cases these posts contain an affective content that is needed to explain the meaning of the review, and that is ignored by recent opinion mining techniques. In this work, we propose an affective annotation model of shared reviews, which takes into account the affectivity expressed by the Web community. The annotation model is then used for improving opinion mining results. A case study demonstrating the effectiveness of the approach is also provided. The method can be exploited to better organize the collaborative work of reviews management, and to better take advantage of reviews inside the community.