Performance Analysis of Deep Approaches on Airbnb Sentiment Reviews

Abstract

Consumer reviews in the Airbnb marketplace are one of the key attributes to measure the quality of services and the main determinant of consumer rentals decisions. Such feedback can impact both a new and repeated consumer's choice decision. The way to manage poor reviews can help to save or damage the host's reputation. Sentiment analysis enables an Airbnb host to get an insight into the business, pinpoint degradation of the specific component of compound services and assist in managing it proactively. Multiple Deep Learning algorithms have been used for Natural Language Processing (NLP). For optimal sentiment management in the Airbnb marketplace, it is crucial to identify the right algorithm. The paper uses multiple Deep Learning algorithms to identify different aspects of guest reviews and analyze their accuracies. The paper uses four accuracy measurement benchmarks Precision, Recall, F1-score and Support to analyze results. The analysis shows that the GRU method achieves the best results with the highest classification metrics values as compared to RNN and LSTM.