Extraction of Polarity from Textual Information based on Machine Learning Approach

Abstract

Many of the major corporations use the public platform to advertise their products and services. As a result, the feedback given by their numerous customers is the foundation of their popularity among the general public. Manually categorizing the feedback into positive polarity or negative polarity can be a huge time consumable. The classification process is fully automated to improve results. Since reviews are inherently unstructured, they should be pre-processed and POS-labelled before being used to categorize human polarity. Term Frequencies are used to extract and weigh the terms that are mainly sentiment bearing (TF). The reviews are classified using supervised Machine Learning algorithms. The working of the algorithms for classification is evaluated using performance parameters such as Support terms, Term Frequency, Inverse document frequency, weight vector, and opinion power.