Utilizing Deep Learning for Aspect-Based Sentiment Analysis in Restaurant Reviews

Consumers rely on social media opinions to make product choices and purchases. With the popularity of web-based platforms like Tripadvisor, consumers express their opinions and feelings about food quality, service, and other aspects affecting restaurants through comments. Hence, analyzing these comments can be valuable for others to choose a restaurant or to improve and develop their products and brands. Sentiment analysis utilizes text mining methods to extract, identify, and study emotions and subjective perceptions. Since consumers can use comments to choose a restaurant, this study seeks to provide sentiment analysis of their reviews on the Tripadvisor website about Iranian restaurants. This study is applied in nature, aiming to analyze and manually label 4000 comments from the Tripadvisor website regarding restaurants in ten tourist cities across Iran. It uses a standard extended long short-term memory algorithm for sentiment analysis, a deep learning neural network, and Python text mining packages for modeling. The results indicate that the F-Measure for all aspects exceeds 80%, indicating sufficient efficiency and accuracy of the aspect-based sentiment analysis model for restaurant reviews. The most significant features for customers of Iranian restaurants are the food and the atmosphere. This study represents one of the initial efforts to analyze comments posted on the Tripadvisor website concerning Iranian restaurants. Business owners in the tourism industry, especially restaurant owners, can use the proposed model to automatically and quickly analyze customer feedback, improve performance, and gain a competitive edge. The proposed model can also assist users of online platforms in analyzing the opinions of others, enabling them to make informed decisions more efficiently.