Purpose: The purpose of this study is to propose a citation recommendation system that leverages deep learning techniques to enhance the research process by providing relevant citation suggestions. Design/methodology/approach: This research introduces a supervised deep learning model combining Bidirectional Gated Recurrent Unit (Bi-GRU) with Convolutional Neural Network (CNN) architectures to capture semantic relationships. Additionally, Word2Vec embeddings trained on scientific texts are incorporated to improve the model’s representational capabilities. Citation relationships are used as training labels for text similarity classification. Findings: The proposed model uses citation probability as a metric of technical similarity between papers, outperforming traditional methods like TF-IDF and Doc2Vec in a quantum technology dataset. It excels at providing highly relevant and irrelevant recommendations but faces limitations in suggesting papers with intermediate relevance. Research limitations: The model’s performance in recommending papers with intermediate relevance needs further refinement to improve its overall capacity in handling such cases. Practical implications: This model’s architecture and similarity metric can be integrated into academic search engines, addressing the issue of information overload and enhancing the efficiency of scholarly research. Originality/value: This study contributes to the field of citation recommendation by introducing a novel deep learning-based approach that improves the precision and recall of citation suggestions, offering theoretical advancements and practical solutions for academic research.
Wang et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: