Abstract:

Purposes To clarify the copyright risks triggered by text and data mining （TDM） and large language models （LLM） technologies in the commercial reuse of full-text big data in scientific journals， and to propose actionable compliance solutions based on this analysis. Methods A literature review and case analysis were conducted， issueed by focus group interviews， to extensively collect issues and cases concerning the reuse of scientific journal copyrights in the era of big data. Differences between China’s Copyright Law and the copyright legislation of major global economies （the United States， Japan， the European Union， and the United Kingdom） regarding TDM and LLM usage were examined. The study analyzes the copyright risks associated with the reuse of full-text big data in Chinese scientific journals and discuss and generate compliance authorization recommendations for journals to address potential risks. Findings In light of domestic scientific journal practices， five major copyright risks in the reuse of full-text big data are identified. Against the backdrop of open science， we recommend that journals and authors give priority to “copyright licensing （exclusive/non-exclusive）” over traditional “copyright assignment”clauses. When entering into licensing agreements with databases， journals must ensure they have legally obtained the necessary copyright authorization from authors and clearly define the scope of authorization in the agreement to prevent unauthorized sublicensing to databases beyond the granted rights. We recommend renegotiating copyright-licence agreements among authors， journals， and full-text databases， so that a clear and compliant authorization chain of “authors-journals-databases” should be established. At the same time， we recommend that the government take the lead in establishing a national-level green open storage platform similar to PubMed Central， to provide the necessary infrastructure for the compliant reuse of big data. Conclusions Journals should re-sign copyright agreements with authors and databases that comply with the reuse of full-text big data， improve the rules for journal data security and copyright usage， and encourage scholars to make their research findings more openly accessible. This study provides strategies for balancing copyright protection and data reuse in the era of big data for scientific journals， proposing a set of practical and feasible compliance measures. These measures ensure that journals fully utilize big data resources while respecting authors’ rights， promoting the open sharing and innovative application of academic information.

Key words: Journal copyright, Text and data mining, Large language model, AI-generated content, Big data reuse

摘要：

目的 厘清大数据时代文本与数据挖掘（TDM）与大语言模型（LLM）技术在科技期刊商业性全文大数据再利用场景中所引发著作权风险，并据此提出可操作的合规授权策略。 方法 采用文献复习法、案例分析法，并结合焦点小组访谈，广泛收集大数据时代科技期刊著作权再利用问题和案例，对比我国《中华人民共和国著作权法》与世界主要经济体（美国、日本、欧盟、英国）版权法在TDM/LLM著作权使用政策的差异，分析我国科技期刊全文大数据再利用存在的著作权风险，讨论并产生期刊如何应对潜在风险的合规授权建议。 结果 结合国内科技期刊的实践，指出当前全文大数据再利用中的5项著作权风险。在开放科学背景下，建议期刊与作者优先采用“著作权许可（专有/非专有）”以替代传统的“著作权转让”条款。建议期刊与数据库签订许可协议时，必须确保期刊已合法取得作者的著作权授权，并在协议中清晰界定授权范围，以避免超出授权权限向数据库进行非法转授权。通过期刊与作者及全文数据库重新签订著作权许可协议，明确构建“作者-期刊-数据库”的合规授权链条。同时，建议政府主导建立类似 PubMed Central 的国家级绿色开放仓储平台，为大数据的合规再利用提供必要的基础设施。 结论 期刊须与作者和数据库重新签署符合全文大数据再利用的著作权协议，完善期刊数据安全和著作权使用规则，鼓励学者将科研成果进一步开放获取。本研究为科技期刊在大数据时代的著作权保护与数据再利用之间提供了平衡的策略，提出了一套切实可行的合规授权方案。在确保期刊尊重作者权利的同时，能够充分利用大数据资源，推动学术信息的开放共享与创新应用。

关键词: 期刊著作权, 文本与数据挖掘, 大语言模型, 人工智能生成内容, 大数据再利用