基于LDA2Vec-BERT的新兴技术主题多维指标识别与演化分析研究OA北大核心CHSSCDCSSCICSTPCD

[Purpose/Significance]Mining and visualizing the global disruptive technology,such as the emerging and hot technical topics and their evolution differences implied in the global blockchain patent literature can provide for practi-tioners in the field,science and technology policy makers,management departments and science and technology research and development personnel.[Method/Process]Based on the patent literature in the global blockchain field,different time slices were divided in a timely order.LDA topic model,Word2vec word vector model,and BERT language model were comprehensively utilized to build the technical topic mining model in the blockchain field.At the same time,four-dimension-al indicators,including topic popularity,topic population,topic technicality and topic novelty were constructed to identify the hot technology topics and emerging technology topics in the field of blockchain.These were then combined with the topic evolution model to carry out evolutionary analysis on the emerging hot topics.[Result/Conclusion]The research finds that the LDA2Vec-BERT topic recognition and evolution model could identify emerging and hot technical topics in the field based on the massive patent literature in the field of blockchain,and visually and clearly displays the evolution trend and characteristics of the topic of blockchain technology,and finds the development trend of blockchain technology from archi-tecture research to application research.By comparing the model results,it can be found that the empirical results are rea-sonable,and the accuracy rate,recall rate and F1 value of the model are higher than other recognition models,which proves that the method has the good effect on the recognition of disruptive technology topics.