中国科技期刊知识服务技术路径探析——以Consensus.app为例

Exploring the technical path of knowledge service in Chinese STM journals:A case study of Consensus.app

【目的】通过深入剖析新型知识服务平台Consensus.app的运行模式和技术实施方式,探索大语言模型为中国科技期刊知识服务发展带来的新启示。【方法】以文献调研与网络实证调研相结合的形式,分析Consensus.app平台的功能、特点及知识服务的表现形式,揭示该平台运行的技术实现方式,评估其在促进科技文献知识获取方面的作用,总结其应用中可能存在的优劣性。【结果】Consensus.app的语料库通过SemanticScholar数据库的2亿篇摘要信息构建,该平台运用多种人工智能技术,从研究论文中提取关键信息构建向量化的知识数据库,然后使用OpenAI的接口将用户提问转化为检索语句对知识库进行检索,再基于检索命中信息生成扼要结论反馈给用户。Consensus.app可针对不同问题提供直接基于数据支持的结论,并快速阅读文献快照信息,协助用户快速决策。【结论】Consensus.app可在一定程度上解决大语言模型答案所缺乏的准确性和证据链问题,为科技期刊在大模型时代的广泛高效应用提供更丰富的使用场景,展示结合大语言模型构造知识库来提供科技期刊知识服务的新路径。在新时期,科技期刊界要高度重视数据质量建设,重视跨界合作、版权完善,积极拥抱大模型时代的趋势变化,迎接科技期刊的“AI+”时代。

[Purposes]This paper aims to explore new pathways and insights for knowledge services in Chinese STM journals in the era of large language models by analyzing the operational model and technical implementation of the novel knowledge service platform Consensus.app.[Methods]By combining literature review and online empirical research,this study analyzed the functions,features,and manifestations of knowledge services of the Consensus.app platform.In this paper,the technical implementation methods employed by the platform and its role in promoting knowledge acquisition from scientific literature were investigated,with a summary of its potential advantages and disadvantages.[Findings]The corpus of the Consensus.app platform,based on the abstract information from the Semantic Scholar paper database,employs various artificial intelligence technologies,including natural language processing,machine learning,and information retrieval.By extracting key information from research papers and creating a vectorized knowledge database,Consensus.app utilizes OpenAI′s interface to retrieve relevant information from the knowledge base based on user queries and provides summarized conclusions as feedback to users.The platform offers highly personalized interactions via direct data-supported conclusions for different queries and quick access to snapshot information of relevant literature,to help users make rapid decisions.[Conclusions]Consensus.app partially addresses the lack of accuracy and evidence chain in large language model responses.It also provides more diverse scenarios for the widespread and efficient application of scientific journals in the era of large language models.It demonstrates a new approach to integrate large language models into knowledge repositories for further knowledge services for STM journals.In the new era,the STM journal community needs to attach great importance to data quality development,cross-disciplinary collaboration,and copyright improvements,and it also needs to embrace the trend in the era of large models to move towards the “AI+”era of STM journals.