Sentiment Analysis of Customer Reviews of Food Delivery Services Using Deep Learning and Explainable Artificial...

Area of Science:

• Computer Science
• Artificial Intelligence
• Business Analytics

Background:

• The COVID-19 pandemic accelerated the growth of food delivery services (FDSs).
• Customer reviews on online platforms are crucial for FDS performance evaluation and improvement.
• FDS organizations seek to leverage customer feedback for enhanced satisfaction.

Purpose of the Study:

• To review machine learning (ML) and deep learning (DL) models for predicting customer sentiments in the FDS domain.
• To explore the role of explainable artificial intelligence (XAI) methods in conjunction with these models.
• To identify current trends and challenges in sentiment analysis within the FDS sector.

Main Methods:

• Systematic literature review of ML, DL, and XAI techniques applied to FDS customer reviews.
• Analysis of existing studies on sentiment prediction models in the food delivery industry.
• Evaluation of model interpretability and explainability in sentiment analysis.

Main Results:

• Lexicon-based and ML techniques are widely used for sentiment analysis in FDS.
• Deep learning (DL) models show promise but lack interpretability, hindering adoption.
• A significant majority (77%) of reviewed models are non-interpretable, impacting trust and explainability.

Conclusions:

• Implementing DL models for sentiment analysis in FDS is recommended.
• Integrating XAI techniques is crucial to enhance model explainability and build organizational trust.
• Future research should prioritize explainable DL models for FDS sentiment analysis to improve customer satisfaction.