Sentiment Analysis of User-Generated Data Using CNN-BiLSTM Model

With the availability of user-generated data, like reviews, it has become possible to gain valuable information that can help service providers to improve their services. There can be various methods to gain helpful information, including sentiment analysis of the reviews. Sentiment analysis is a popular method used to extract information from these reviews, but it can be challenging due to irregularities and ambiguities in the data. This chapter proposes a novel model architecture that enhances sentiment analysis and provides more accurate results than state-of-the-art technology. The proposed model takes advantage of the convolutional layer and Bidirectional long short-term layer (BiLSTM) coupled with max-pooling and a network of hidden layers. Regularizers are also incorporated to enhance the accuracy of the proposed model during training. The proposed model’s main advantage is its ability to provide more accurate sentiment analysis results compared to previous models. In our chapter, we observed that the model overcomes the limitations of earlier models by improving the handling of irregularities and ambiguities in unstructured data (user-generated reviews) and providing valuable insights to service providers by accurately classifying reviews into positive and negative categories, enabling them to take appropriate actions based on customer feedback.