Detecting Sentiments from Movie Reviews by Integrating Reviewers Own Prejudice
Presently, sentiment analysis algorithms are widely used to extract positive or negative feedback scores of various objects on the basis of the text/reviews. But, an individual may have a certain degree of biasness towards a certain product/company and hence may not objectively review the object. We try to combat this biasness problem by incorporating the positive and negative bias component in the existing sentiment score of the object. This paper proposes several algorithms for a new system of implementing individual bias in the corpus of data i.e. movie reviews in this case. Each review comment has an unadjusted sentiment score associated with it. This unadjusted score is refined to give an adjusted score using the positive and negative bias score. The bias score is calculated using certain parameters, the weightage of which has been determined by conducting a survey. We lay emphasis on the degree of biasness an individual has towards or against the review parameters for the movie reviews corpus namely actor, director and genre. We equip the system with the capability to handle various scenarios like positive inclination of the user, negative inclination of the user, presence of both positive and negative inclination of the user and neutral attitude of the user by implementing the formulae we developed.
