An AI-driven Restaurant Recommendation Application
Wentian (Manu) Zhu, School of Computing, Computer science, 1st year undergraduate
Dharma Tejaswini Janga, School of Computing, Cyber security and privacy, 1st year undergraduate
Chase MacMillan, School of Computing, Young Dawgs Internship
The application is designed to provide users with personalized dining suggestions based on their cuisine preferences, price range, distance, and meal type (Snack, Light Meal, or Full Meal). Leveraging generative AI technologies, the system efficiently extracts and analyzes restaurant reviews, filters out irrelevant information, and presents users with meaningful insights to aid their decision-making. The platform dynamically updates recommendations based on user interactions and ensures a streamlined experience by avoiding duplicate suggestions within a single session. Additionally, the system integrates with mapping services to provide real-time travel distances and directions.
System Features and Functionality
Our system offers a range of features to enhance the restaurant discovery experience. It intelligently filters and sorts restaurants based on user-defined preferences, ensuring that recommendations align with individual tastes and needs. One of the key functionalities is the AI-powered review summarization, which processes customer feedback from various sources and extracts relevant insights. This allows users to quickly understand the strengths and weaknesses of a restaurant without having to sift through long reviews. Additionally, the system utilizes an AI-enhanced multi-dimensional rating system to provide more detailed assessments. Rather than relying on a single overall score, our system generates ratings in categories such as food quality, service, ambiance,
cleanliness, and overall experience, offering a more nuanced evaluation.
Use of Generative AI Technologies
Generative AI plays a crucial role in enhancing multiple aspects of the recommendation system. One of its primary applications is in the pre-processing and summarization of restaurant reviews. Many online reviews are embedded in complex HTML structures that include extraneous elements such as icons, code snippets, and unrelated numerical values. To extract meaningful content, we use advanced AI models such as GPT-4o and Llama 3.3, which identify and isolate key review components, including the reviewer’s name, publication date, rating, and main review text. The extracted information is structured into a standardized format (e.g., JSON), ensuring efficient processing and integration.
User Experience and Example Output
The system is designed to provide a seamless, intuitive, and intelligent dining recommendation experience. Each restaurant recommendation includes a summary of the restaurant’s cuisine type, customer sentiment analysis, an AI-generated rating breakdown, the estimated distance from the user, and a Google Maps link for easy navigation. Additionally, if a restaurant has fewer than five reviews, the system acknowledges the limited data and encourages users to contribute by trying out new places.
Conclusion and Future Enhancements
By integrating generative AI at multiple stages of the recommendation process, our system improves the accuracy, personalization, and efficiency of restaurant discovery. Users receive context-aware, data-driven, and highly relevant dining suggestions that enhance their overall experience.
Looking ahead, we plan to implement additional features, including user review contributions, which will allow AI to continuously learn from user feedback and further refine recommendations. Additionally, image analysis technology will be introduced to assess menu images and customer-uploaded photos, providing users with visual insights into food quality and presentation. Finally, the integration of voice-based search capabilities will make the restaurant discovery process even more seamless and accessible.