Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Panoramic Interests: Stylistic-Content Aware Personalized Headline Generation

Junhong Lian, Xiang Ao, Xinyu Liu, Yang Liu, Qing He

TL;DR

SCAPE tackles the gap in personalized headline generation by modeling users' panoramic interests that encompass both content preferences and headline stylistics. It introduces a three-component framework: a headline inference module that uses LLMs to extract content and style features from headlines, a hierarchical fusion network that combines long- and short-term interests for both content and style, and a personalized injection mechanism that guides a lightweight generator. A disentanglement strategy with contrastive loss prevents entanglement of content and style signals, improving stability and interpretability. Experiments on the PENS dataset show SCAPE achieves state-of-the-art performance in informativeness, fluency, and personalization, demonstrating the practical value of incorporating stylistic-content aware user representations into headline generation. The work highlights the importance of panoramic user modeling for content-aware and style-aware personalization in news headlines, with potential implications for broader text generation tasks.

Abstract

Personalized news headline generation aims to provide users with attention-grabbing headlines that are tailored to their preferences. Prevailing methods focus on user-oriented content preferences, but most of them overlook the fact that diverse stylistic preferences are integral to users' panoramic interests, leading to suboptimal personalization. In view of this, we propose a novel Stylistic-Content Aware Personalized Headline Generation (SCAPE) framework. SCAPE extracts both content and stylistic features from headlines with the aid of large language model (LLM) collaboration. It further adaptively integrates users' long- and short-term interests through a contrastive learning-based hierarchical fusion network. By incorporating the panoramic interests into the headline generator, SCAPE reflects users' stylistic-content preferences during the generation process. Extensive experiments on the real-world dataset PENS demonstrate the superiority of SCAPE over baselines.

Panoramic Interests: Stylistic-Content Aware Personalized Headline Generation

TL;DR

SCAPE tackles the gap in personalized headline generation by modeling users' panoramic interests that encompass both content preferences and headline stylistics. It introduces a three-component framework: a headline inference module that uses LLMs to extract content and style features from headlines, a hierarchical fusion network that combines long- and short-term interests for both content and style, and a personalized injection mechanism that guides a lightweight generator. A disentanglement strategy with contrastive loss prevents entanglement of content and style signals, improving stability and interpretability. Experiments on the PENS dataset show SCAPE achieves state-of-the-art performance in informativeness, fluency, and personalization, demonstrating the practical value of incorporating stylistic-content aware user representations into headline generation. The work highlights the importance of panoramic user modeling for content-aware and style-aware personalization in news headlines, with potential implications for broader text generation tasks.

Abstract

Personalized news headline generation aims to provide users with attention-grabbing headlines that are tailored to their preferences. Prevailing methods focus on user-oriented content preferences, but most of them overlook the fact that diverse stylistic preferences are integral to users' panoramic interests, leading to suboptimal personalization. In view of this, we propose a novel Stylistic-Content Aware Personalized Headline Generation (SCAPE) framework. SCAPE extracts both content and stylistic features from headlines with the aid of large language model (LLM) collaboration. It further adaptively integrates users' long- and short-term interests through a contrastive learning-based hierarchical fusion network. By incorporating the panoramic interests into the headline generator, SCAPE reflects users' stylistic-content preferences during the generation process. Extensive experiments on the real-world dataset PENS demonstrate the superiority of SCAPE over baselines.
Paper Structure (15 sections, 10 equations, 3 figures, 1 table)

This paper contains 15 sections, 10 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Formulation
  4. Our SCAPE Framework
  5. Headline Inference
  6. Hierarchical Stylistic-Content Awareness Fusion
  7. Personalized Injection Module Guided Generator
  8. Disentanglement Strategy
  9. Experiments
  10. Experimental Setup
  11. Datasets and Baselines
  12. Evaluation Metrics
  13. Implementation Details
  14. Results and Analysis
  15. Conclusion

Figures (3)

  • Figure 1: Illustration of the Joint Influence of Content Interests and Stylistic Preferences on Headline Personalization.
  • Figure 2: Architecture of the Proposed SCAPE Framework.
  • Figure 3: Win Rates in Personalization Evaluation.