Modeling Self-Representation Label Correlations for Textual Aspects and Emojis Recommendation.

Area of Science:

• Computer Science
• Information Retrieval
• Artificial Intelligence

Background:

• The proliferation of online platforms necessitates advanced recommender systems.
• User reviews, encompassing text and emojis, offer rich data for aspect-based recommendations.
• Existing systems struggle to effectively model personal preferences and semantic correlations within review aspects.

Purpose of the Study:

• To propose a novel correlation-aware review aspect recommender (CARAR) system.
• To enhance personalized recommendations by integrating textual and emoji aspects.
• To improve the handling of personal preferences and semantic correlations in recommendations.

Main Methods:

• Developed the CARAR model to construct self-representation correlations between textual and emoji aspects.
• Utilized identified dependencies between textual aspects and emojis to learn user and item latent factors.
• Incorporated cross-view correlation mapping and real-world environmental information to refine recommendations.

Main Results:

• The CARAR model demonstrated superior performance compared to six existing models across five datasets.
• Experimental results validated the effectiveness of the proposed approach in review aspect recommendation.
• The model successfully captured and leveraged cross-view correlations for enhanced recommendation quality.

Conclusions:

• The CARAR system offers a significant advancement in review aspect recommendation.
• Integrating multi-modal review aspects (text and emoji) improves recommendation accuracy.
• The model's ability to capture complex correlations leads to more personalized and effective suggestions.