ELSA: Ensemble learning based sentiment analysis for diversified text

Abstract

Internet-based media and online business make people use it a lot, especially the high use of cell phones everywhere in the world. These encourage organizations to gain insights from people far away. This creates a tremendous growth of information on the Internet. Concepts are important implications for practically all human practices and practices. It looks for feelings before making a decision. People share their opinions, thoughts, ideas, attitudes, feelings etc. with known languages on social media platform. Most people express their opinions in mixed languages (English with their mother tongue). Analyzing diverse texts is a very challenging area for practitioners and researchers. In this research work, it presents an ensemble technique by collaborating Generative Adversarial network (GAN) and Self-Attention Network (SAN) to analyze the tweets that contains diversified languages. The GAN layer helps to arrange the preprocessed tweets into positive and negative. The SAN layer helps to identify the neutral tweets. The proposed technique is compared along with the existing Concurrent Neural Network (CNN) along with Self-Attention Network (SAN) on Tanglish tweets. The proposed work produced better accuracy than the existing work.