基于LDA2Vec和DTM模型的颠覆性技术主题识别研究——以能源科技领域为例

Identification of Disruptive Technology Topics Based on LDA2Vec and DTM Models:A Case Study in the Energy Technology Field

[目的/意义]颠覆性技术关乎国家竞争力和国际地位,科学准确地识别出颠覆性技术主题,能够解决技术发展过程中主题不够明确、发展路径不够清晰等问题,以此有效把握技术发展动态,调整国家科技战略布局,更好地抢占国际竞争制高点。[方法/过程]以能源科技领域的专利文本数据为研究客体,构建基于Word2Vec词向量与LDA(Latent Dirichlet Allocation)主题向量的融合特征向量,并引入K-means算法优化主题聚类效果,最后结合颠覆性技术特征指标,识别颠覆性技术主题,利用DTM(Dynamic Topic Model)模型揭示该领域颠覆性技术主题的发展状况。[结果/结论]通过人工验证和模型结果对比可以发现,实证结果具有合理性,且模型的精准率、召回率、F1值均高于同类型的主题模型,证明该方法对颠覆性技术主题识别具有较好效果。

[Purpose/Significance]Disruptive technology is crucial to a country's competitiveness and inter-national status.Accurately identifying the topics of disruptive technology can effectively address the problems of unclear themes and development paths in the process of technology development.This can provide a better grasp of the dynamics of technology development,adjust the layout of national science and technology strategies,and better seize the international competitive high ground.[Method/Process]This study used patent text data in the field of energy technology as the research object,constructed a fusion feature vector based on Word2Vec word vectors and LDA topic vectors,and introduced the K-means algorithm to optimize the topic clustering effect.Finally,combin-ing with the characteristic indicators of disruptive technology,the disruptive technology topics were identified,and the DTM model was used to reveal the development status of disruptive technology topics in this field.[Result/Conclusion]Through manual verification and comparison with model results,the empirical results are reasonable.The precision,recall,and F1 values of the model are higher than those of similar topic models,which proves that this method is effective in identifying disruptive technology topics.