Abstract:
Sentiment negation and negation scoring can be considered as major aspects of sentiment analysis. Social media sentiment analysis can be considered as an excellent source of information in today's business. But there is very minimal work has been done in sentiment negation scoring. All the existing negation scoring mechanisms are based on an adjective intensity approach. This research proposes a novel mechanism for negation scoring based on an adjective-adverb combined approach. The domain is movie reviews which are obtained from international movie database. We have introduced new negation axioms that all negation scoring mechanisms must satisfy. Three new negation scoring functions which satisfy those negation axioms have been introduced here to calculate word, sentence and document level sentiment negation. These are all based on adjective-adverb hybrid approach. We describe our experimental results on different 700 training and different 300 testing movie reviews. We compare our functions with existing sentiment negation functions. It can be seen that our functions perform better than the existing functions based on Pearson correlation with respect to human subjects. It can be concluded that these new functions perform an overall accuracy of 71% in precision.
