<![CDATA[Scientific literature review is a fundamental process in academic research that requires significant time and effort. This study proposes a novel framework that combines Retrieval-Augmented Generation (RAG) with Citation-Enhanced mechanisms and a Multi-Factor Ranking algorithm to automate the scientific literature review process intelligently and accurately. The proposed approach integrates three main components: (1) semantic-based document retrieval module using dense vector embeddings, (2) a citation augmentation system that analyzes citation networks between scientific papers, and (3) a multi-factor ranking algorithm that considers semantic relevance, citation impact, publication recency, and author authority. Experiments were conducted on the S2ORC (Semantic Scholar Open Research Corpus) dataset containing over 200,000 scientific papers across various domains. Evaluation using ROUGE-L, BLEU-4, BERTScore, and Citation F1 metrics demonstrates that the proposed approach yields significant improvements over conventional RAG methods. The proposed system achieves a ROUGE-L score of 0.612 and BERTScore of 0.847, improving by 8.3% and 6.1% respectively compared to standard RAG baseline. The results demonstrate that integrating citation information in the retrieval and text generation process substantially enhances the quality, accuracy, and completeness of automatically generated literature reviews.]]>
Copyrights © 2026
