基于语义相似度的技术多主题演化路径识别方法研究

Multi-thematic Evolution of Technology Based on Semantic Similarity

[目的/意义]基于技术演化路径识别方法研究的现状,提出一种基于语义相似度的有向有权网络技术多主题演化路径识别方法,有助于企业依据自身优势和目标定位,选择研发方向和实现路径。[方法/过程]利用文本挖掘中的主题模型将专利文本向量化,计算具有引用关系专利之间的主题相似度,构造有向有权网络,并利用Java编写的遍历算法识别出涵盖特定技术领域的多主题演化路径,供企业把握领域内技术主题的发展变化。[结果/结论]以太阳能电池板为例的研究结果表明,该方法可以有效地识别出太阳能电池板技术围绕5个主题在不断演进的路径。与传统单一路径识别方法相比,本研究提出的基于语义的多路径识别,不仅弥补了传统单一路径不能准确揭示技术关联主题的局限,也为企业依据自身技术优势选择研发方向提供了决策依据。

[Purpose/Significance]Based on the status quo of the research on the technology evolution path identification method,this paper proposes a multi-thematic evolution path identification method based on semantic similarity for directed-to-ownership network technology,which helps enterprises to choose the direction of R D and the realization path according to their own advantages and goals.[Method/Process]By using the topic model in text mining to vectorize the patent texts,calculating the topic similarity between the patents with reference and constructing the directed and authorized network,and use the traversal algorithm written in JAVA to identify the multi-thematic evolution paths covering specific technical fields,the reserch allows enterprises to grasp the development and changes in the field of theme technology.[Result/Conclusion]The case study of solar panels shows that the method proposed can effectively identify the evolution path of solar panels technology that functions around the five topics in the. Compared to the traditional single-path identification method,the multi-path recognition based on semantics not only makes up for the limitation of the traditional single-path in failing to accurately reveal the topic of technology association,but also provides the decision-making basis for enterprises to choose the R D direction based on their own technological advantages.