Abstract:

Purposes Maintain an open attitude towards the rapidly evolving generative artificial intelligence （GenAI）， apply it to multiple stages of English scientific journal publishing， preserve the integrity of academic literature， and improve the publication efficiency. Methods Seven AI tools were selected and tested in our routine publishing workflows （Jianziyuan， Grammarly， FigCheck， AJE， DeepEdit， ChunlinAI-Editor， and Doubao）. Their performance was evaluated in terms of machine-generated text detection， image similarity checking， English language polishing， English manuscript editing， Chinese abstract revision， and promotional material generation/refinement. Findings For machine-generated text detection， the test results showed a relatively high false-positive rate； for image similarity checking， the tested AI tool could effectively detect problematic images； for English language polishing， the tested AI tools showed effectiveness； for English manuscript editing， the tested AI tool demonstrated high efficiency in identifying English spelling/grammar issues， as well as specific formatting ones； for Chinese abstract revision， the tested AI tool showed key advantage in the accurate translation of domain-specific terms； for promotional material generation/refinement， the tested AI tool showed high efficiency and good user-friendliness. Conclusions AI tools can effectively empower academic publishing. Currently， all outputs generated by AI tools strongly undergo final manual review by editors. We adopt a human-centered， gradual advancement strategy： in areas where AI excels， fully leverage its capabilities； in areas where AI shows potential， we stay tuned and proceed with cautious verification.

Key words: Generative artificial intelligence （GenAI）, Academic publishing, Empowerment evaluation, Journal editing workflow

摘要：

目的 对发展速度惊人的生成式人工智能持开放态度，将其运用于英文科技期刊出版各环节，维护学术文献的完整性，提高编辑出版效率。 方法 选取7种AI工具（鉴字源、Grammarly、FigCheck、AJE、DeepEdit、ChunlinAI-Editor和豆包），结合编辑部日常出版流程各环节，测试这些工具在识别机器生成文本、图片相似度检测、英文润色、英文稿件编辑校对、中文摘要修改、宣传文本素材生成/完善等方面的实际效果。 结果 所测试的AI工具识别机器生成文本误报率较高；能有效识别问题图片；英文润色总体效果不错；针对编辑校对阶段的英文拼写语法/部分格式识别效率高；中文摘要修改学术专业名词贴合度较好；高效生成/完善宣传文本素材。 结论 AI工具有效赋能学术出版。现阶段，所有AI工具的处理结果，都必须依赖编辑进行最终的人工把关。采取以人为本，渐进式前行的策略——在AI擅长的领域，最大程度发挥其效能；在AI有潜力的领域，保持关注并谨慎验证。

关键词: 生成式人工智能, 学术出版, 赋能评估, 期刊编辑工作流