摘要
[目的/意义]技术创新机会识别能够帮助企业更好地实现技术布局战略,从而抢占技术制高点以形成科技创新的非对称优势。[方法/过程]文章提出了一种基于生成式拓扑映射(The Generative Topographic Mapping, GTM)的技术创新机会识别方法。结合深度学习、机器学习等方法从SCI论文及德温特专利数据入手进行技术创新机会识别。首先构建基于Bert-LSTM文本分类模型,实现对论文及专利数据的主题识别;然后运用自然语言处理(Natural Language Processing, NLP)技术中词性标注对论文和专利摘要文本进行挖掘,从而识别出不同研究领域的关键技术词;最后采用GTM算法绘制技术地图,通过对空白新兴技术点逆向映射进而实现对技术创新机会的有效识别。[结果/结论]以新能源汽车为例进行验证,结果显示基于Bert-LSTM文本分类模型的预测精度优于其他分类模型;基于论文和专利数据绘制的技术地图,能够直接呈现出不同数据源所涉及的研究范围,通过GTM算法逆向映射实现了快速、客观识别技术创新机会,从而为企业技术布局提供数据支撑。本研究为客观识别技术创新机会提供了新的解决思路,以期为相关研究者提供有益启示。
[Purpose/significance] Technological innovation opportunity identification can help enterprises to better realize the technology layout strategy, therefore to seize the technological high ground to form the asymmetric advantage of scientific and technological innovation.[Method/process] In this paper, an approach is offered to identify innovation opportunities based on The Generative Topographic Mapping(GTM).And the deep learning and machine learning methods are applied to identify technological innovation opportunities predicated on SCI papers and Derwent patent data.Firstly, we build a classification model based on Bert-LSTM to identify the topics of papers and patent data.Secondly, we use the natural language processing(NLP) technology to mine the abstracts of papers and patents to identify the key technology words in different research fields.Finally, we apply the GTM algorithm to draw a technology map, thus identifying the technology innovation opportunities effectively through the reverse mapping of the blank emerging technology dots.[Result/conclusion] This paper takes new energy vehicles as an example for validation, and the results show that the prediction accuracy of the Bert-LSTM text-based classification model is superior to other classification models.Further more, the technology map based on the papers and patent data can directly present the scope of research involved in different data sources, and the GTM algorithm reverse mapping enables rapid and objective identification of technology innovation opportunities, thus providing data support for enterprise technology layout.Ultimately, this study offers a new solution for identifying technology innovation opportunities objectively, with a view to providing useful insights for relevant researchers.
出处
《情报理论与实践》
CSSCI
北大核心
2021年第6期146-153,198,共9页
Information Studies:Theory & Application
基金
国家自然科学基金青年项目“关键利益相关者视角下新兴产业创新政策作用机制与仿真优化:以新能源汽车为例”的研究成果,项目编号:71704101。
关键词
数据挖掘
深度学习
技术创新机会
技术地图
主题识别
data mining
deep learning
technological innovation opportunities
technology map
topic identification
