MVP: Memetic Voter Patterns for aspect extraction in sentiment analysis

The reviews in online resources are used extensively to evaluate the quality of various subjects, such as products. The products have several aspects, and extracting these aspects from review texts is a sub-task of aspect-based sentiment analysis. This paper proposes a novel algorithm, named Memetic Voter Patterns, or MVP, to identify aspect words, by using patterns of parts-of-speeches of their adjacent words. This method yields a higher accuracy than previous methods and gives a better understanding of the structure of sentences around an aspect. This method is based on a voting system. The patterns of the part-of-speeches around aspects are identified by incorporating a memetic algorithm. This method can have applications in other subfields of opinion mining, such as finding sentiment phrases.