Towards Seamless Air Travel: Developing a Next-Generation Flight Booking Assistant

Abstract

In the dynamic landscape of flight booking, passengers seek an optimal and streamlined experience encompassing factors such as flight duration, onboard services, and pricing. This paper introduces developing a Next-Generation Flight Booking Assistant, leveraging Meta’s LLAMA 2, a transformer-based auto-regressive causal language model. Our approach not only provides a comprehensive solution to the traditional challenges of flight booking but introduces several novel features. We fine-tune the LLAMA 2 model using the Low-Rank Adaptation (LoRA) technique,  enabling efficient handling of user queries.  Furthermore,  we introduce a destination recommendation model based on travel history, employing the Retrieval-Augmented Generator (RAG) for data retrieval on booking and pricing information. The Next-Generation Flight Booking Assistant incorporates distinctive features, including real-time interpretation of social media sentiment to grasp user preferences, seam- less handling of multiple languages, and the provision of post-booking recommendations for top places to visit. These distinguishing aspects set our assistant apart by  actively adapting to the evolving needs of users. To justify the significance of our solution, we present comprehensive performance statistics derived from user testing and feedback collection. The evaluation underscores the assistant’s proficiency in understanding user preferences, navigating through available options, and efficiently completing the booking process. The results demonstrate not only the effectiveness but also the user satisfaction with the introduced features, solidifying the Next-Generation Flight Booking Assistant as a cutting-edge and user-centric solution in the realm of digital flight assistance.