Research On Identification Of Emerging Cross-Domain Based On Patent Analysis

The solutions to recent scientific problems often involve multiple fields,and the crossfertilization of knowledge is an important trend in current technological development and an important way for emerging technologies to emerge.Effective and timely identification of emerging cross-cutting areas has important guiding significance for national science and technology policy formulation and strategic layout of science and technology.In recent years,the research on the identification of emerging cross-cutting areas has continued to increase,among which the bibliometric-based method is more widely used,but this method only performs statistical analysis on the external characteristics of scientific and technical literature,and lacks semantic analysis on the content level of the literature.To address this problem,this paper takes the two dimensions of interdisciplinarity and emergence as the entry point,combines the semantic information of patent texts with citation relations,predicts the disciplinary classification of patents using text classification algorithms in deep learning,determines the interdisciplinary nature of patents,and identifies emerging technologies from the perspective of disciplinary cross-fertilization.The main research work of this paper includes the following four parts:(1)A method for identifying emerging cross-cutting fields based on semantic information of patents is proposed.In this method,first,the semantic information in patent titles and abstracts is mined by BERT pre-training model;then,a multi-label classification model is trained for identifying cross-domain patents in the patent dataset;finally,the subject matter of the technology domain is obtained through the cross-domain patents,and the patent index method is used to discern whether the technology domain is an emerging cross-domain.(2)To further improve the accuracy of the multi-label classification model,a multi-label classification model integrating patent semantic information and patent citation information is proposed.The experimental results show that the sample features that can be extracted by relying only on the semantic information of patents are limited,and the accuracy of the fused multi-feature classification model is improved by 6.19% compared with that of the semantic feature-based classification model.(3)An empirical study of the identification method proposed in this study is conducted in the intersection of quantum physics and computer technology.In the experimental results,the emerging intersection fields such as "quantum communication" and "quantum computing" are identified.Compared with the existing cross-cutting identification methods,the proposed method can identify emerging cross-cutting areas that are still in their infancy or growth stage.(4)The emerging cross-field identification system is designed and implemented to realize the functions of patent data analysis and emerging cross-field identification,providing a patent analysis platform for professionals to grasp the law of technology development.The test results of the system show that the emerging cross-field identification system realized in this paper has certain reliability.