CPJN: News recommendation with a content and popularity joint network

Area of Science:

• Computer Science
• Information Retrieval
• Artificial Intelligence

Background:

• News recommendation systems often overlook the dual factors influencing user clicks: content interest and news popularity.
• Existing models primarily focus on content preferences, neglecting in-depth analysis of popularity-driven engagement and independent preference modeling.

Purpose of the Study:

• To propose a novel Content and Popularity Joint Network (CPJN) model for enhanced news recommendation.
• To independently model user preferences for news content and popularity, addressing limitations in existing approaches.

Main Methods:

• Developed a CPJN model comprising three interconnected networks: content-based, popularity-based, and adaptive combination.
• The content-based network eliminates popularity bias from user side information to refine content preference representation.
• The popularity-based network removes content bias from news side information to model popularity preference.

Main Results:

• The CPJN model demonstrated significant effectiveness on two real-world datasets.
• Achieved average improvements of 1.493% in accuracy rate and 1.502% in recall rate compared to state-of-the-art baselines.

Conclusions:

• The proposed CPJN model successfully integrates user content and popularity preferences for superior news recommendation.
• The adaptive combination network effectively handles varying user sensitivities to news popularity, further boosting recommendation performance.