Sentiment Analysis through Text using Squeeze and Excitation Attention Mechanism based Recurrent Neural Network

Sentiment analysis is the field of research to identify helpful data by utilizing sentiments of the people shared through social media like Twitter and Facebook. However, some sentiment analysis approaches attain minimum accuracy due to inconsistencies between the tweet text and the assigned sentiment label. Hence, this research proposes the Squeeze and Excitation Attention mechanism based Recurrent Neural Network (SEA-RNN) approach for the sentimental analysis in the text. Initially, the Twitter United States (US) Airline dataset is performed to implement the proposed method. Then, pre-processing techniques like removing the stop words, stemming and lemmatization are employed to enhance the quality of the dataset. Then, Bidirectional Encoder Representations from Transformers (BERT) is employed in the feature extraction process to identify and extract relevant features. Finally, the SEA-RNN approach is proposed for classifying sentiments into three classes such as positive, neutral and negative. The experiment results show that the proposed SEA-RNN attains a better accuracy of 99.84% as compared to the existing methods like Universal Language Model Fine-Tuning" (ULMFiT) with Support Vector Machine (SVM).