A Novel Hybrid Integration of BERT and Conventional Machine Learning Techniques for Robust Airline Twitter Sentiment Analysis

The rapid expansion of social media applications like Twitter has led to an influx of user-generated content, presenting vast opportunities for sentiment analysis. In this paper, a novel approach is proposed that combines the power of BERT, a DL model, with conventional ML algorithms for sentiment analysis with airline Twitter data. The method uses BERT’s contextual understanding of natural language to encode tweets into dense representations, capturing nuanced sentiment information. Later, several conventional ML algorithms applied (SVC, Logistic Regressions, and Random Forest) in conjunction with BERT to further enhance sentiment analysis accuracy to 96.5%. Our experimental results on a Twitter sentiment dataset demonstrate that the hybrid approach significantly improves sentiment classification performance compared to using BERT or conventional algorithms alone. The proposed method provided comprehensive evaluations, including classification reports and accuracy scores, to validate the effectiveness of our proposed approach.