Smart Schemes Recommender System: A Systematic Review of AI-Based Approaches for Government Welfare Scheme Personalization

Authors:

Somasekhar T¹*, Lohith PC², Mahadev K³, Manish M⁴, Anil Kumar N⁵

¹ Associate Professor, Department of CSE, KSIT, Karnataka, India

²–⁵ Students, Department of CSE, KSIT, Karnataka, India

*Corresponding Author: Somasekhar T

Abstract - Governments worldwide launch hundreds of welfare schemes annually across education, healthcare, agriculture, and employment. Despite their availability, many eligible beneficiaries remain unaware due to inefficient dissemination channels, complex eligibility criteria, and language barriers. This systematic review surveys artificial intelligence (AI), machine learning (ML), and natural language processing (NLP) approaches applied to government scheme recommendation systems. Thirty studies published between 2023 and 2026 are reviewed to identify dominant techniques, research trends, performance benchmarks, and critical research gaps. Findings reveal that hybrid recommendation models combining content-based filtering and collaborative filtering consistently outperform single-method approaches, while large language models (LLMs) and retrieval-augmented generation (RAG) represent the most promising emerging directions. Key research gaps include the absence of multilingual support, limited explainability, inadequate privacy mechanisms, and lack of real-time scheme updates. Based on the review, a scalable system architecture is proposed that integrates these techniques to deliver personalized, transparent, and accessible welfare scheme recommendations. The proposed system aims to enhance digital governance, reduce citizen exclusion, and improve the utilization of public welfare programs in developing countries such as India.

Key Words: Government Scheme Recommendation, Machine Learning, Natural Language Processing, Hybrid Filtering, E-Governance, Recommender System.