A time-robust group recommender for featured comments on news platforms

Area of Science:

• Computer Science
• Information Retrieval
• Natural Language Processing

Background:

• Content moderators face challenges in selecting high-quality comments for news platforms due to manual processes and platform growth.
• Automated systems are needed to assist moderators in efficiently identifying and featuring valuable user comments.

Discussion:

• Expert evaluation highlights the subjective nature of comment moderation, supporting the utility of comment recommendation systems.
• The research emphasizes the value of recommending comments over simple classification for aiding human moderators.

Key Insights:

• A group recommender system integrating comment data, user history, and contextual relevance achieves high ranking scores.
• The best-performing models maintain ranking performance (mean NDCG@5 of 0.89) across changing article topics and platform growth.
• Realistic evaluation scenarios using unseen online discussions from different years validate the models' effectiveness.

Outlook:

• This work advances semi-automated content moderation techniques.
• Further research can explore enhancing deliberation quality assessment in online discourse through improved recommendation algorithms.