Polarity Analysis Based on an Improved Feature Selection Algorithm

Polarity analysis is a technique that analyzes text according to the attitude implied from it. Since the correspondence between the attitude and surface word features is subtle, the ordinary methods for topical text analysis are not sufficient for polarity analysis. This paper proposed a classification method combined with feature selection algorithm to the problem. Firstly, all possible feature schemas are designed and features are collected. Then a feature selection algorithm named improved SIMBA is used to distill the candidate features. In the last, the selected features are used as clues for a standard classifier. Experiment shows that the performance is better than that of previous method on the same dataset.