摘要：

【目的】目前大多数审稿平台已推出了审稿人智能遴选系统,但现有基于传统机器学习的方法对特征的提取依赖人工规划,且非线性特征学习能力不足,难以捕捉稿件与审稿人之间的潜在关联,导致稿件与审稿人的匹配度有限。【方法】将生成对抗网络模型融入到审稿人智能遴选算法,通过自动学习提取复杂特征,捕捉传统方法难以识别的潜在关联,动态调整匹配策略,以实现更加精准高效的审稿人匹配机制。【结果】将审稿人智能遴选算法嵌入到北京勤云科技的远程稿件处理系统8.0中,并对其进行验证和对比分析,结果表明,嵌入本文所提算法的系统表现出较高的推荐准确率,且推荐准确率明显优于对比方法,尤其在技术性较强的领域及跨学科稿件场景中表现尤为突出。【结论】基于生成对抗网络的审稿人智能遴选算法为审稿人推荐提供了一种更准确、高效的方法。

关键词: 审稿人智能遴选, 生成对抗网络, 匹配策略, 推荐准确率

Abstract:

[Purposes] Most current peer review platforms have implemented intelligent reviewer selection systems. However, existing methods based on traditional machine learning rely on manual feature engineering and exhibit limited nonlinear feature learning capabilities, failing to capture latent associations between manuscripts and reviewers, which restricts the accuracy of manuscript-reviewer matching. [Methods] This study integrates a generative adversarial network (GAN) model into the intelligent reviewer selection algorithm. By automatically learning and extracting complex features, the proposed method identifies potential correlations overlooked by traditional approaches and dynamically adjusts matching strategies to achieve a more precise and efficient reviewer-manuscript matching mechanism. [Findings] The algorithm was embedded into Beijing Qinyun Technology’s remote manuscript processing system (version 8.0) for validation and comparative analysis. Results demonstrate that the system incorporating the proposed algorithm achieves significantly higher recommendation accuracy compared to baseline methods, particularly excelling in technically specialized fields and interdisciplinary manuscript scenarios. [Conclusions] The GAN-based intelligent reviewer selection algorithm provides a more accurate and efficient method for reviewer recommendation, offering a robust solution to enhance the quality and fairness of academic peer review.

Key words: Intelligent reviewer selection, Generative adversarial network (GAN), Dynamic matching strategies, Recommendation accuracy