Sentiment Analysis to Quantify Healthcare Data

Abstract

Every organization that interacts with some form of customers or users has a feedback system. Feedback provides a vital source of information to the organization about how the end users feel about their service. However, textual feedbacks are very subjective, and to be able to use them for rating or mathematical analysis, we need to quantify the textual feedback. Sentiment analysis can be performed on the textual feedback to attain a quantifiable output. In this paper we aim to design two systems; one that learns from raw text examples by clustering (unsupervised learning) the text samples and then assigning classes to these clusters, and another system that uses this trained data and classifiers to classify new textual data into one of the sentiment classes. For clustering we use K-means clustering method and discuss the performance of the same, and for classification we use two classifiers; K-nearest neighbors (KNN) and Naïve Bayes (NB). Finally, we compare the performance of the two proposed classifiers.