科技情报基础服务知识融合流程自动化框架研究

Automation Framework Study for Knowledge Fusion of Technology Intelligence Basic Services

科技情报的知识融合可以为科技情报分析提供高质量、高效率的基础数据。在知识融合自动化过程中,快速使用和整合现有技术,推进语义层面的知识融合,是科技情报快速感知和问题解决的关键。本文通过梳理科技情报基础数据服务中多层次、多粒度服务对知识融合的需求,设计了基于模块化、分阶段、任务管理式的自动化融合框架。框架可以快速组合基于规则的方法、基于监督学习的方法、大语言模型方法等,对多种来源的科技情报数据进行抽取、转化、合并、知识生成和关系拓展。最后在商业竞争对手跟踪场景中实践了基于该框架的新闻事件提取和关联。结果证明,基于模块化、自动化方式的科技情报融合能够更高效地应用各类方法的优点,为科技情报融合难点问题、流程化问题提供更多元的解决方案。

Knowledge fusion of scientific and technological intelligence(S&T intelligence)can provide high-quality and efficient basic data for the analysis of scientific and technological intelligence.In the process of knowledge fusion automation,rapid use and integration of existing technologies to advance knowledge fusion at the semantic level is the key to rapid perception and problem solving of scientific and technological intelligence.In this paper,by sorting out the demand for knowledge fusion in multi-level and multi-granularity services in the basic services of scientific and technological intelligence,we design an automated fusion framework based on a modularised,phased,task-managed approach,which can quickly combine rule-based methods,supervised learning-based methods,and large language model methods,etc.,for extracting,transforming,merging,generating knowledge,and expanding relationships of scientific and technological intelligence data from a wide range of sources.Finally,we practice this approach in a business competitor tracking scenario to extract event from news,and form new relationships.The results prove that the automated approach to S&T intelligence fusion can apply the advantages of various methods more efficiently and provide more diversified solution options for the difficult problems of S&T intelligence fusion.