Leveraging Bidirectional LSTM for Enhanced Sentiment Analysis: A Deep Learning Approach

Abstract

In this work, we use LSTM for this model to show that it is effective for NL processing tasks such as sentiment analysis as discussed herein. It was thus the purpose of this paper to leverage LSTM in capturing long-term dependencies of sequential data for improved identifications of sentiments in a big Twitter data set. The rationale for doing so is that the given set of data provided a huge number of textual inputs for training and testing of the proposed model as all the additional tweets were marked as positive, negative, or neutral emotion. Hyper parameters were also adjusted to set the best scenario since we_sess2rnted on believes that such probing areas like several LSTM units, batch size, learning rate, and dropout rates among others could best deliver. As predicted the better-proposed model and the newly proposed improved model got 84% of accuracy this confirms the efficiency of LSTM in sentiment analysis. This method brought out the seriousness of choosing the right model parameter and it also helped proclaim the flexibility of the deep learning model in handling Real-World noisier Social Media data for sentiment analysis.