Sentiment Analysis of Customer Reviews of Food Delivery Services Using Deep Learning and Explainable Artificial Intelligence: Systematic Review

Abstract

During the COVID-19 crisis, customers’ preference in having food delivered to their doorstep instead of waiting in a restaurant has propelled the growth of food delivery services (FDSs). With all restaurants going online and bringing FDSs onboard, such as UberEATS, Menulog or Deliveroo, customer reviews on online platforms have become an important source of information about the company’s performance. FDS organisations aim to gather complaints from customer feedback and effectively use the data to determine the areas for improvement to enhance customer satisfaction. This work aimed to review machine learning (ML) and deep learning (DL) models and explainable artificial intelligence (XAI) methods to predict customer sentiments in the FDS domain. A literature review revealed the wide usage of lexicon-based and ML techniques for predicting sentiments through customer reviews in FDS. However, limited studies applying DL techniques were found due to the lack of the model interpretability and explainability of the decisions made. The key findings of this systematic review are as follows: 77% of the models are non-interpretable in nature, and organisations can argue for the explainability and trust in the system. DL models in other domains perform well in terms of accuracy but lack explainability, which can be achieved with XAI implementation. Future research should focus on implementing DL models for sentiment analysis in the FDS domain and incorporating XAI techniques to bring out the explainability of the models.

Unboxing Deep Learning Model of Food Delivery Service Reviews Using Explainable Artificial Intelligence (XAI) Technique

“…Over the past couple of decades, there has been a good amount of research published on the application of ML techniques to perform sentiment analysis in the FDS domain [ 13 , 28 , 29 , 30 , 31 ]. Sentiment analysis of customer reviews from tweets for various FDSs, such as Swiggy, Zomato and Uber Eats, was performed to understand consumer satisfaction [ 32 ].…”

“…In another significant work, Noor [ 33 ] compared the results of Lexicon, SVM, Natural Language Processing (NLP) and Text Mining from different works and found that Lexicon achieved the highest accuracy of 87.33% compared to other methods. However, it would be difficult to perform a sentiment analysis in languages other than English [ 13 ]. Additionally, domain adaptation must be taken into account while creating models because a word in one domain may have a different meaning in another.…”

“…Additionally, domain adaptation must be taken into account while creating models because a word in one domain may have a different meaning in another. For example, ‘lightweight’ is a positive sentiment word for the electronics domain, whereas it is a negative sentiment word for kitchen appliances [ 13 ]. ML/DL techniques can overcome the challenge of domain adaptation by training the model from the same domain dataset.…”

“…In the context of movies, pictures and movies are synonymous, but they are not in the context of cameras, where they refer to two different things. Furthermore, the terms “photo” and “picture” are synonymous in the camera industry [ 13 ]. Traditional dictionary-based lexicon training approaches do not work well as they are strictly limited to a smaller number of words, whereas ML/DL techniques can overcome these problems [ 34 ].…”

“…Customers primarily look for reviews and recommendations of others when they order cuisines from online FDSs. Positive reviews drive customers to make a decision on ordering food from one restaurant, whereas negative reviews help them look for other options [ 13 ]. FDS companies can look for the negative comments towards common complaint types, such as customer service, food quality, cost and slow delivery service, to understand the improvement areas to enhance customer satisfaction.…”