Optimizing restaurant recommendations through sentiment analysis

Abstract

"The rapid growth of the food industry and dining-out culture has led to a vast growth in the number of restaurants over the years. As a result, customers are overwhelmed with too many options to purchase their meals from. Using sentiment analysis, this study’s goal would therefore be to create a data-driven recommendation system that can provide customers with the best restaurants that serve their preferred cuisine. Unlike previous research, this study includes a model that analyzes reviews in four languages namely English, Bangla, Code Switch (Bangla and English) and Code Mix (Banglish) from a multitude of food and applications. A customer’s choice of restaurant will depend on multiple factors such as the recommendations of friends or food critics, their culinary preference, pricing, location, the general reputation of the restaurant and so on. This study proposes a novel approach that uses sentiment anal- ysis based on primarily sourced restaurant reviews and ratings to provide person- alized restaurant recommendations to interested customers. The assessment of our proposed model will be wide-ranging, through sentiment analysis applying multiple BERT model variants such as BERT for English annotated reviews, BanglaBERT for Bengali reviews, BanglishBERT for Code-Mixed reviews and XLM-RoBERTa for Code-Switched reviews with 79% , 74%, 75% and 77% best model accuracies respectively. Furthermore, this research investigates the analysis of various LSTM architectures, including Attention-LSTM, Bi-LSTM, and a Transformer-based T5 model. Utilizing customer reviews for a variety of restaurants, the efficiency of the model across multiple languages and types of sentiment analysis tasks will be eval- uated throughout the entire set of experiments. The results of this study should provide a time efficient solution for food enthusiasts and the general consumer to be introduced to the finest restaurants with the best gastronomic delights, magnificent ambience and atmosphere and the best customer service. "