In the past three decades, the competitive airline industry has experienced rapid growth. However, choosing a particular flight for travelling remains a crucial task and requires detail analysis and prior experience. These could be leveraged by intellectually processing existing passengers’ reviews or feedbacks. Traditional methods of collecting and analyzing such feedbacks or reviews are time consuming and repetitive. To address this challenge, sentiment analysis can be an effective tool. Twitter data has emerged as a valuable resource for collecting such reviews or feedbacks and conducting sentiment analysis. This research paper introduces a new deep learning model that combines BERT (a language model), self-attention based Bidirectional Gated Recurrent Unit with Capsule network, and Softmax classifier to assess airline passenger review tweets on six different US airlines, with three sentiment classes: positive, neutral, and negative. The performance of the sentiment analysis system is evaluated with various contemporary methods. This assessment demonstrates that the suggested model exhibits comparable performance to state-of-the-art methods.