Visual Recommendation for Peer-To-Peer Accommodation with Online Reviews based on Sentiment Analysis and Topic Models

Area of Science:

• Artificial Intelligence
• Data Science
• Consumer Behavior Analysis

Background:

• The sharing economy, particularly peer-to-peer (P2P) accommodation, is rapidly growing.
• User-generated content (reviews) in P2P accommodation is rich in insights but overwhelming in volume.
• Effective analysis of reviews is crucial for consumer decision-making and service improvement.

Purpose of the Study:

• To develop a comprehensive pipeline for personalized accommodation recommendations.
• To extract actionable insights from user reviews for both consumers and managers.
• To create a visual analytic system for interactive exploration of accommodation data.

Main Methods:

• Topic modeling to mine opinions from user reviews.
• Deep learning for sentiment analysis of reviews.
• Aspect-level sentiment analysis to generate sentiment vectors for accommodations.
• Development of a personalized recommendation algorithm.
• Design of a user-friendly visual analytic system.

Main Results:

• A robust pipeline for processing P2P accommodation reviews.
• Generation of sentiment vectors representing accommodation characteristics.
• A personalized accommodation recommendation system.
• A visual analytic system facilitating interactive data exploration.
• Demonstrated usefulness and effectiveness through user and case studies.

Conclusions:

• The proposed method effectively processes user reviews for personalized recommendations.
• The system provides valuable insights for both consumers and P2P accommodation managers.
• Visual analytics enhance the understanding and usability of review data.