Aspect Ranking Based on Author Specific Information Aggregation

Abstract

In the present huge volumes of reviews about entities are available in the websites. Hence the necessity to summarize the opinions and to understand about the polarity of the features (aspects) in the review has become very essential. Opinion summarization helps in summarizing the overall opinions of all the aspects in the review. However, extracting the aspects from the review is critical because the writing style of the authors changes from person to person and it is also a challenging task. The existing system specifies that the aspects exist in the review in the form of explicit aspect expressions. Aspect expressions occur frequently as noun and noun phrases in the review. Aspects are extracted by the use of feature dependency parsing algorithm in the domain specific ontology. Opinion summarization deals with the polarity aggregation (positive or negative) on each of these aspects. The proposed approach summarizes the importance of aspects using the author specific preference aggregation model. This approach considers the preferences of authors for each aspect in addition to various measures like frequency, weighted score along with the polarity. These preferences are aggregated in the ontology using the method of pair wise comparison. This paper aims at integrating all the scores which will help to extract the aspects specific to the domain and summarizes the aspects based on its rank in an efficient manner.