作为推动社会进步与经济发展的重要力量,如何快速且精确地识别潜在颠覆性技术对于提升企业竞争力、推动产业变革等具有关键意义.本文提出一种融合专利数据与社交媒体数据的潜在颠覆性技术早期识别方法,是对已有研究理论和方法的重要补充.第一,检索相关领域专利数据,对数据进行划分;第二,基于颠覆性技术的特征,选取与颠覆性技术相关的指标以构建指标体系,并计算其技术影响力;第三,依托Bi-LSTM(bi-directional long short-term memory)训练专利指标与技术影响力之间的关系,预测出候选颠覆性技术,并结合BERTopic提取技术主题;第四,通过BERTopic主题建模基于社交媒体数据提取出社会主题,并通过关注度和情感倾向对社会主题进行评价;第五,通过语义相似度,将社会主题与技术主题匹配映射,以对技术主题进行分类,进而识别出潜在颠覆性技术;第六,以医疗机器人为例,阐述该识别方法的应用过程.研究结果表明,Bi-LSTM模型在准确率、精准率、召回率和F1-score上均超过90%,优于其他模型;将医疗机器人领域中的潜在颠覆性技术划分为高关注度-积极态度、低关注度-积极态度与低关注度-消极态度3种类型;识别出的医疗机器人潜在颠覆性技术,能够为国家发展政策制定与相关产业布局提供参考.
A Deep Learning Approach for Identification of Potentially Disruptive Technologies by Integrating Patent Data and Social Media
In promoting social progress and economic development,prompt and accurate identification of potentially dis-ruptive technologies is critical for enhancing enterprise competitiveness and driving industrial transformation.This study proposes a method for the early identification of potentially disruptive technologies by integrating patent and social media data,as an important supplement to existing research theories and methods.First,the relevant patent data in the field were retrieved and divided.Second,based on the characteristics of disruptive technologies,indicators related to disruptive tech-nologies were selected to construct an indicator system and calculate the technological impact.Third,Bi-LSTM was em-ployed to train the relationship between patent indicators and technological impact,whereby candidate disruptive technolo-gies were predicted and combined with BERTopic to extract technology topics.Social topics were extracted from social media data through BERTopic modeling and evaluated for attention and sentiment tendencies.Subsequently,using seman-tic similarity,the social topics were matched and mapped to technology topics for classification,to identify potentially dis-ruptive technologies.The application of this identification method is illustrated in medical robots.The results demonstrated that the Bi-LSTM model achieves above 90%accuracy,precision,recall,and F1-score metrics,outperforming other mod-els.Potentially disruptive technologies in the field of medical robotics can be classified into high attention-positive atti-tude,low attention-positive attitude,and low attention-negative attitude.The potentially disruptive technologies identified in medical robotics can provide valuable references for national development policies and relevant industry arrangements.
potentially disruptive technologiespatent datasocial media dataBi-LSTMmedical robots
万方数据
维普
