NLP-Based Sentiment Analysis using Deep Learning Methods
Dheeraj Kumar, Sandeep Bhatia, Hardeep Singh Dhillon, Amit Kumar Goel · 2024
Opinion mining systems rely heavily on sentiment analysis because to the vast amount of data and opinions that are generated, exchanged, and sent on a regular basis through the internet and other media. This study presents a deep learning network-based sentiment analysis categorization that was created and compares the results across multiple deep learning networks. Using the Multilayer Perceptron, a benchmark for other networks' performance was established (MLP). Built and implemented on the 50K movie review file IMDB dataset were a hybrid model of long short-term memory (LSTM) and convolutional neural network (CNN), as well as an LSTM recurrent neural network. Reviews for the dataset were 50% positive and 50% negative. After Word2Vec performed its initial pre-processing on the data, word embedding was applied. The hybrid CNN_LSTM model outperformed the MLP and the separate CNN and LSTM networks, as seen by the findings. While CNN reported an accuracy of 88.9%, MLP and LSTM reported accuracy of 87.78% and 87.68, respectively. 89.4% accuracy was reported by CNN_LSTM. Results further show that the proposed deep learning models have performed better than SVM, Naïve Bayes, and RNTN models based on English datasets presented in previous works.