Analyzing the Sentiments with Neural Network
Niharika Niharika, Sona Malhotra · Journal of scientific research · 2021
In the modern era, ubiquitous means of communications are used.Thus, allowing various platforms where people can exchange their thoughts and opinions.This leads to the situation where the task of sentiment analysis comes into play which is the task of natural language processing.Many approaches such as Maximum Entropy, classifier of Naïve Bayes were used for the purpose of analyzing the sentiments of the users which helps the internet surfers to interpret whether a particular thing is liked or disliked by others.The paper proposes a system with backpropagation neural network (BPNN) which not only analyzes simple reviews but also the opinions that contain emoticons and negation.The neural network was implemented with various datasets and with several different activation functions such as ReLu (Rectified Linear Unit), sigmoid and hyperbolic tangent.The results depict that sigmoid activation function performed better than other.