New framework improves explainability for recommender systems

Researchers have developed a new framework for explaining recommender systems, addressing scalability issues with existing post-hoc methods. This approach uses spectral biclustering to group users and items, allowing entire blocks of interactions to be removed for analysis. Evaluations on the MovieLens and Amazon datasets using Singular Value Decomposition and Neural Collaborative Filtering models demonstrate that specific interaction blocks significantly impact recommendation quality, with user segments showing varying sensitivities to these removals. AI

IMPACT This research offers a more scalable and practical approach to understanding how recommender systems function, potentially increasing trust and transparency for users.

RANK_REASON This is a research paper detailing a new framework for explainability in recommender systems. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• MovieLens
• ScienceCast
• scite Smart Citations
• Singular Value Decomposition

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →