[目的/意义]文章对科技政策隐性扩散路径自组织方法进行研究，挖掘科技政策文本包含深层语义信息， 将隐性知识显性化，为科研人员拓展和丰富政策扩散路径研究提供参考。[方法/过程]本文结合科技政策篇章文 本的形式语义和内容语义两个方面对政策文本结构化处理和深度挖掘，对政策文本资源全解析，抽取科技政策文 本中包含的特征，其中包括概念和关系自动获取与标引技术、网络表示学习，挖掘科技政策文本中的隐含结构信 息，利用BiLSTM-CRF模型的深度学习方法实现概念的自动获取和自动标引关系。将得到多篇科技政策文本的 概念和关系组成概念关系对的形式，借助于表示学习的方法发现每个节点稠密的向量表示。[结果/结论]通过实 验验证，证明了本文借助隐性路径特征的科技政策扩散隐性路径自组织方法的有效性，在一定程度上拓展了政策 研究的方法，为科研人员在政策扩散研究上提供了参考。
[Purpose/significance] This paper studies the self-organization method of the recessive diffusion path of science and technology policy, excavates the deep semantic information contained in the text of science and technology policy, makes the tacit knowledge explicit, and provides reference for researchers to expand and enrich the research of policy diffusion path. [Method/process] This paper combines the formal semantics and content semantics of science and technology policy texts to deal with the structural processing and deep mining of policy texts, fully analyzes the poli? cy text resources, and extracts the features contained in the science and technology policy texts, including the automat? ic acquisition and indexing technology of concepts and relationships, network representation learning, mining the im? plicit structural information in the science and technology policy texts, and using the deep learning method of BiLSTMCRF model to realize the automatic acquisition and indexing of concepts. The concepts and relationships of many sci? ence and technology policy texts are formed into concept relationship pairs, and dense vector representations of each node are found by means of representation learning. [Result/conclusion] Through the experimental verification, it is proved that the self- organization method of the hidden path of science and technology policy diffusion based on the characteristics of the hidden path in this paper is effective, which expands the method of policy research to a certain ex? tent and provides a reference for researchers in policy diffusion research.
