基于论文和专利资源的技术机会发现研究

Scientific paper and patent data are two largest information resources in the world. The technology opportunities with vitality and potential commercial value usually leave some footprints in these two information resources. The paper and patent information resources are often considered as the indicators of basic science studies and technologies, respectively. Intuitively, there should be some interactive and exclusive relationships between paper and patent information resources, and it can be advantageous to analyze these two kinds of information resources simultaneously. Therefore, it is very important for the integration analysis of paper and patent information resources to grasp technology opportunities, understand the relationship between science and technology, and improve the level of technology innovation.. On the basis of existing S&T information resource processing procedure, this project will discover in deep the underlying relationship between the technology topics for research and invention from respective paper and patent information resources. The authors and research institutions in paper information resource will be linked to the inventors and applicants in patent information resources. Meanwhile, the citation relationships between paper and patent information resources are also constructed. In this way, one can change the lonely island characteristics (that is to say, locally order but globally disorder) of existing information resources. . This project will study some key common technologies on mining deeply multiple information resources with different structures, and provide some supporting toolbox for domestic community. Specifically, this project will focus on the life-cycle model of technology and dynamic technology topic analysis, and then detect technology opportunities on the basis of integrated information resources. In the end, the resulting S&T information obtained from this project can help our country develop industrial policy, the main body of technological innovation grasp the specific direction of R&D investment, find some suitable strategic partners and so on.

论文和专利资源是描述科研产出的两大主要表现形式，具有生命力和潜在商业价值的技术机会在这两大信息源中通常会留下一定的足迹。论文信息资源通常被用于测度基础科学研究活动的水平，而专利信息资源则被用于测度产业技术的创新水平，因此论文和专利信息资源的融合分析对于把握技术机会、理解科学和技术之间的联系、提高技术创新水平具有重要意义。. 本课题在现有科技信息资源加工的基础上，基于内容深入揭示发明主题与研究主题间的关系，建立作者与发明人、研究机构与申请人、论文和专利间的引用等的关联，改变现有信息资源局部有序但整体无序的孤岛特征；攻克深度挖掘利用多源异构信息资源的关键共性技术并形成相应的工具支撑，以技术生命周期阶段判别和技术主题演化分析为方法，在信息资源融合的基础上发现潜在的技术机会，为国家制定产业政策、为创新主体把握特定技术研发投资方向、寻找合适的战略合作伙伴等提供科技情报支撑。

论文和专利资源是描述科研产出的两大主要表现形式，具有生命力和潜在商业价值的技术机会在这两大信息源中通常会留下一定的足迹。论文信息资源通常被用于测度基础科学研究活动的水平，而专利信息资源则被用于测度产业技术的创新水平，因此论文和专利信息资源的融合分析对于把握技术机会、理解科学和技术之间的联系、提高技术创新水平具有重要意义。.本课题在现有科技信息资源加工的基础上，基于内容深入揭示发明主题与研究主题间的关系，建立作者与发明人、研究机构与申请人、论文和专利间的引用等的关联，改变现有信息资源局部有序但整体无序的孤岛特征；攻克深度挖掘利用多源异构信息资源的关键共性技术并形成相应的工具支撑，以技术生命周期阶段判别和技术主题演化分析为方法，在信息资源融合的基础上发现潜在的技术机会，为国家制定产业政策、为创新主体把握特定技术研发投资方向、寻找合适的战略合作伙伴等提供科技情报支撑。.本课题主要研究进展有以下四个方面：（1）提出了一种主题关联强度分析方法，从而可揭示论文和专利资源间存在的相互影响和排斥关系，并透视科学和技术之间的发展规律；（2）技术术语间的语义关系对成功预测特定领域的技术机会起到关键作用，受Rel-LDA、Type-LDA以及TNG模型的启发，提出了Rel-TNG和Type-TNG模型，有机融合了多元语法特征；（3）多学科交叉使得技术主题结构缺乏明确的边界，技术主题识别问题被看作从大规模引文网络中检测重叠社区的问题，混合成份随机块模型被用于检测重叠社区结构，随机变分推断算法用于推断模型参数，有效解决了大规模引文网络技术主题结构检测的问题；（4）基于本项目以及本项目负责人主持的其他课题的研究成果，带领团队成功研发了大数据环境下多源信息融合的科技文献智能分析服务平台SciTeMiner，有机嵌入了技术主题分析和技术机会发现模块。