| 姜芳,夏那日朵兰.智能大模型在科技期刊论文初校中的应用研究与建议——以《信息安全学报》为例.编辑学报,2026,38(3):341-346 |
| 智能大模型在科技期刊论文初校中的应用研究与建议——以《信息安全学报》为例 |
| Research and recommendations on the application of artificial intelligence in the initial editorial review of sci-tech journal papers:a case study of the Journal of Cyber Security |
| |
| DOI：10.16811/j.cnki.1001-4314.2026.03.016 |
| 中文关键词: 智能大模型 科技期刊 论文初校 人机协同 |
| 英文关键词: large language models sci-tech journals initial manuscript review human-machine collaboration |
| 基金项目: |
|
| 摘要点击次数: 158 |
| 全文下载次数: 27 |
| 中文摘要: |
| 本文以《信息安全学报》出版的200篇网络空间安全论文为样本,构建5个维度10项分析指标,对比分析智能初校与人工初校的结果,总结得出智能初校在校对效率、标准化格式核查、基础规范校验方面优势突出,但存在幻觉错误、非结构化信息识别薄弱、缺乏专业逻辑研判能力等问题；人工校对更擅长把控专业细节、学术逻辑与内容真伪,却存在效率低、标准化不均的短板,二者能力高度互补。因此本文建议从建立专业规则库、搭建专用智能体、规范大模型提示词模板3个方面提升人工智能初校能力。在此基础上,构建智能大模型与人工结合的科技期刊初校全流程体系,为期刊初校数字化、智能化转型提供思路借鉴与实践参考。 |
| 英文摘要: |
| This study takes 200 published cyberspace security papers from the Journal of Cyber Security as research samples and establishes 10 analytical indicators covering five dimensions to conduct a comparative analysis of preliminary proofreading outcomes generated by AI-powered intelligent proofreading and manual proofreading. The results indicate that intelligent preliminary proofreading boasts prominent advantages in proofreading efficiency, standardized format inspection and basic specification verification, yet suffers from inherent drawbacks including AI hallucination errors, poor recognition of unstructured information and insufficient capacity for professional logical assessment. By contrast, manual proofreading excels at refining professional details, auditing academic logic and verifying content authenticity, but is hampered by low working efficiency and inconsistent standardization across operators; the two proofreading modes feature highly complementary strengths and weaknesses. Accordingly, this paper proposes three targeted strategies to improve the preliminary proofreading performance of AI:developing a domain-specific rule database, constructing dedicated domain AI agents, and standardizing prompt templates for large language models. On this basis, a full-process preliminary proofreading framework integrating large intelligent models and human reviewers is established, offering actionable ideas and practical references for the digital and intelligent transformation of preliminary proofreading in sci-tech journals. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
|
|
