Cellular Automaton-Based Sentiment Analysis Using Deep Learning Methods

The swift expansion of internet-centric applications, including social media platforms, online marketplaces and blogs, has given rise to comments, experiences, sentiments, evaluations and reviews including daily life events. Sentiment analysis involves the collection of thoughts, opinions, experiences and impressions on diverse areas, including product reviews on e-market sites, movie reviews and experiences on various e-services. The existing solutions encounter various challenges in effectively managing such diverse and nuanced data. So we present a new approach for classifying such sentiments using the cellular automata–based word vectorization method with deep learning techniques such as long short-term memory (LSTM), convolutional neural network (CNN) and CNN-LSTM models. The designed model makes use of pertinent properties of the signals obtained from the cellular automaton to classify sentiments as positive or negative. We consider a grid of cells governed by a synchronous update based on Wolfram’s rules. The cellular automaton evolution process slowly learns the context of the text and reaches a stable state by generating cycle length (CL) signals, after achieving convergence. These signals are used for creating a CL weight matrix. The CL weight matrix is then back-propagated to classify sentiments using standard deep learning methods. The proposed system is a new automated method for assessing semantic linkages and the meaning of the text for natural language applications. LSTM performs well with CL signal weights and produces an accuracy level of 99.18% and an F1_score of 98.68% when trained with sentiment tweets. The experimental results illustrate that the suggested approach has the potential to accurately classify social media comments into different classes and consistently outperforms other baseline models, achieving overall good results.