Harnessing AI for Customer Feedback: Sentiment Analysis of US Airlines Tweets Using BERT and Deep Learning Models

Sentiment analysis of social media content, particularly tweets, presents a significant challenge due to the informal and concise nature of the text, yet it is crucial for industries such as airlines where customer feedback can directly influence business operations and reputation. Despite the availability of numerous computational techniques, accurately interpreting the nuances of language used in social media remains a challenge. This study explores the application of various computational techniques to classify sentiments expressed in US airline tweets sourced from a comprehensive Kaggle dataset. Addressing this, our study applied and compared various methods: traditional machine learning models, deep learning models (RNN, LSTM, GRU), and a state-of-the-art BERT transformer model, each offering distinct advantages in handling linguistic complexities. Our results demonstrate that while traditional models achieved respectable accuracies up to 0.78 using Logistic Regression and SVM, deep learning models outperformed them slightly with accuracies reaching up to 0.88 using GRU. The BERT model, however, showcased superior performance with an impressive accuracy of 0.93, proving its efficacy in extracting deep contextual meanings from text. This comprehensive evaluation not only highlights the effectiveness of advanced NLP techniques in sentiment analysis but also sets a pathway for their application in enhancing real-time business analytics and customer relationship management.