Sentiment Analysis of Airline Customer Reviews in Vietnamese Language Using Deep Learning

Abstract

Nowadays, most customers read product reviews before buying. Online reviews have become extremely important in customers’ decision-making. In the aviation industry, there are many environments where customers can give reviews, and others can read these reviews. Reading and processing these reviews is often essential because the number of reviews is vast, and customer reviews about a problem often vary. There are studies about sentiment analysis of airline customer reviews in English. To our knowledge, no research has been conducted on the Vietnamese. Therefore, this study analyzes airline customer sentiment reviews in Vietnamese. We use Deep Learning (DL) models and combinations of these models for sentiment analysis. Previous studies in sentiment analysis focused on the content of reviews; this study proposes to combine the review’s title and content. We create a dataset for this task. The Sequential BiLSTM-CNN model, particularly when combined with the PhoBERT embedding method, demonstrated superior performance across all measured metrics, including accuracy and F1-score, 95% and 0.96, respectively.