Sentiment analysis on medical text using combination of machine learning and SO-CAL scoring

Identifying emotional polarization in a medical report is important in screening, acquiring and synthesizing knowledge of physicians before making a clinical decision. We consider this as a classification problem whose input is a set of sentences collected from medical articles and output is the polarization of each sentence labeled as a positive, negative or neutral one. In this paper, we propose to combine machine learning with natural language processing techniques. For machine learning, we use three features, namely, N-gram, Change Phrase, and Negative ones, extracted from a data set to build an emotion-polarization analysis system. Simultaneously, we incorporate SO-CAL scoring into the system. Our experiments show that this combination improves the classification accuracy.