Comparative Analysis of Artificial Neural Network and Recurrent Neural Network for Sentiment Analysis in Movie Reviews

The ability to recognize and extract sentiments from text data depends heavily on sentiment analysis. We use the dataset of IMDb movie reviews in this work to investigate the use of machine learning and deep learning approaches for sentiment analysis. Using the sentiments indicated in the text, the goal is to categorize movie reviews as positive or negative. To extract pertinent features for the machine learning methodology from the movie reviews, we use conventional feature engineering techniques. These characteristics include sentiment lexicons, bag-of-words, and n-grams. Train well-known machine learning algorithms on these features to create sentiment classifiers, including Naive Bayes, Support Vector Machines, and Logistic Regression. The deep learning approach, in contrast, uses the strength of neural networks to automatically uncover representations from unprocessed text data. We use a recurrent neural network (RNN), more specifically an LSTM network, to capture the sequential flow of language and extract contextual information from the movie reviews. The LSTM network is then used to learn sentiment representations on the IMDb movie review dataset. split the data into test and learn sets so we could compare different approaches. We use several metrics, including accuracy, precision, recall, and F1-score, to evaluate the performance of the sentiment classifiers. We also compare the results of the machine learning and deep learning models to identify the benefits and drawbacks of each approach. Different machine learning models were analyzed for their performance, with Linear SVM coming out on top with an accuracy of 89.57%. Other top performers included Multinomial Naive Bayes, Linear SVM, and XGboost.