A framework for diversifying recommendation lists by user interest expansion.

Area of Science:

• Computer Science
• Information Science

Background:

• Recommender systems are crucial for managing information overload by predicting user preferences.
• Traditional algorithms often lack diversity, focusing narrowly on dominant user interests and neglecting broader preferences.
• Existing methods rarely leverage semantic information like tags to improve recommendation diversity.

Purpose of the Study:

• To propose a novel recommender system framework that enhances recommendation diversity.
• To address the limitations of traditional algorithms in generating diverse and comprehensive item lists.
• To exploit social tagging information for expanding user interests and improving recommendation quality.

Main Methods:

• A framework employing an interest expansion strategy based on social tagging information.
• Expanding user-item interaction records in size and category to represent broader interests.
• Integrating the expanded user profiles with traditional recommendation models for list generation.

Main Results:

• The proposed method effectively improves the diversity of item recommendations.
• Empirical evaluations on three real-world datasets demonstrate enhanced recommendation accuracy.
• The framework successfully broadens the coverage of user interests in recommendations.

Conclusions:

• The novel framework significantly enhances both accuracy and diversity in recommender systems.
• Leveraging social tagging information is a viable strategy for improving user interest representation.
• The approach offers a promising solution for generating more comprehensive and satisfying recommendations.