| 刘雨农,石静,梁琴琴.基于STM 的颠覆性技术主题识别研究[J].情报工程,2023,9(3):081-091 |
| 基于STM 的颠覆性技术主题识别研究 |
| STM-based Topic Identification for Disruptive Technologies |
| |
| DOI：10.3772/j.issn.2095-915X.2023.03.007 |
| 中文关键词: 颠覆性技术主题；颠覆性指数；结构主题模型 |
| 英文关键词: Disruptive technology topics; disruptive index; structural topic model |
| 基金项目:国家重点研发计划课题“全球创新主体创新感知系统”（2019YFA0707203）。 |
| 作者 | 单位 | | 刘雨农 | 1. 中国科学技术信息研究所 北京 100038； | | 石静 | 2. 南京大学信息管理学院 南京 210023 | | 梁琴琴 | 1. 中国科学技术信息研究所 北京 100038； |
|
| 摘要点击次数: 558 |
| 全文下载次数: 591 |
| 中文摘要: |
| [ 目的/ 意义] 基于专利数据，探讨识别颠覆性技术主题并揭示主题关联的方法。[ 方法/ 过程] 以人工智能领域为例，获取目标专利及其引用专利与施引专利，以此计算目标专利的颠覆性指数。基于目标专利摘要，建立以颠覆性指数为协变量的结构主题模型。对领域主题进行分类并构建主题关联网络，同时计算主题流行度，筛选出颠覆性技术主题。[ 局限] 无法完全替代领域专家的经验和智慧，对颠覆性技术主题的预测能力相对有限。[ 结果/ 结论] 得到人机交互、量子人工智能、机器阅读理解和推荐系统四个潜在颠覆性技术主题，发现当前人工智能领域的颠覆性技术创新聚焦于降低算力成本和优化人机互动两个方向。 |
| 英文摘要: |
| [Objective/Significance] This paper explores methods for identifying disruptive technology topics and topic associations based on patent data. [Methods/Processes] First, taking artificial intelligence technology as an example, the target patent and its cited patents and cited patents are obtained to calculate the disruptive index of the target patent. Secondly, based on the target patent abstracts, a structural topic model with the disruptive index as a covariate is built. Finally, the domain topics are classified and the association network is constructed, and the topic popularity is also calculated to filter out disruptive technology topics. [Limitations] The methods can’t replace the experience of expert, and is underperforming on disruptive technology topics predicting tasks. [Results/Conclusions] Four potentially disruptive technology topics, human-computer interaction, quantum artificial intelligence, machine reading understanding and recommender systems, were obtained. The results show a focus on two directions: reducing the cost of computing power and optimizing human-machine interaction. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
