面向TRIZ理论使用者的多标签专利分类  

Multi-label Patent Classification Oriented to TRIZ Users

作　　者：袁力[1] 陈阳[1] 赵勇[1] 

机构地区：[1]华中科技大学自动化学院,武汉430074

出　　处：《计算机科学》2013年第11A期255-258,266,共5页Computer Science

基　　金：高等学校博士学科点专项科研基金项目(20100142120088);国家高技术研究发展计划(2009AA04Z107)资助

摘　　要：专利是创新的结果,更是再创造的知识源泉,对专利技术知识依据创新需求的分类可有效帮助设计者进行创新设计。依据TRIZ理论对产品专利进行自动分类,以辅助利用专利蕴含的技术冲突进行产品创新设计。TRIZ原始的发明原理过于抽象以及有些原理之间有重叠,文中对40个原始的发明原理进行重组,形成20个新的类别。专利自动分类是一类典型的多标签分类问题,文中从Pro_Techniques和CREAX两个软件中收集了针对发明原理进行具体解释的专利数据,并依据此数据集对问题转换和自适应算法两类多标签分类算法进行对比分析。采用海明损失、测度等评估特性评估了上述算法的性能和质量。结果表明,在使用TRIZ专利数据集时,问题转换方法分类性能要明显优于自适应算法。Patent is not only the results but also the resources of products innovation. It can help the designer make innovation effectively, if we classify technique knowledge of patent based on innovative demand. The classification of products patent based on the TRIZ can assistant in using technique contradiction addressed in patent to make innovative design. The original Inventive Principles is so abstract that some principles are overlapped. Paper analyzed the 40 IPs and grouped them into new 20 classes. Patent classification problem is known as multi-label classification problem. Pro- Techniques,CREAX, these two softwares supply patents which explain Inventive Principles in detail. The dataset is used to compare the performance of multi-label classification algorithm~ problem transformation and algorithm adaptation. Several measures such as hamming loss, F-measure have been proposed in the literature for the evaluation of multilable classifiers. The result shows Problem Transformation performs more excellent than Algorithm Adaptation using TRIZ patent datasets.

关 键 词：专利分类 发明原理 TRIZ 多标签 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]