Unveiling Ambivalence in Reviews: Using Sentence-BERT and K-NN for Airline Recommendations

Abstract

There are a number of benefits to travel, when it comes to crossing foreign borders. People are free to share opinions by posting reviews on websites or other online platforms. Customer interactions are directly impacted by reviews. These viewpoints may be presented in a single review positive, negative or conflicted. The studies proved that conflicting internet reviews, has gained enormous attention in recent years. In along with assisting consumers to choose an appropriate airline, these ratings also help airline companies find and fix problems with their services. In order to fill this gap, we propose a study that conceptualizes the features of contradictory airline evaluations, identifying the perceptions of travelers that lead to their attitudinal ambiguity or doubt while formulating their actions. We studied to see how travelers' attitudinal ambivalence, which results in indecision, is triggered by contradicting aspects of airline evaluations. In this paper, we proposed many strategies to eliminate this ambivalence and offer suggestions to the customers as well as airlines. We firstly pre-processed traveler evaluations in the recommendation system using NLP (Natural Language Processing) approaches. After pre-processing Sentence-BERT is used to convert the textual reviews into vectors and then K-NN is used to recommend the airline based on those vectors. This machine learning technique suggests a suitable airline to customers. The machine learning model's recommendation accuracy is increased by this novel approach to using online social networks to advertise low-cost flights to travelers. Our proposed model achieved an accuracy of 94.7%, precision of 95.24%, F1 score of 95.4% and recall of 95.5%.