【作者】 高继平；

【导师】 丁堃；

【作者基本信息】 大连理工大学 ， 科学学与科技管理， 2013， 博士

【副题名】以SIPOD中数字信息的传输（HO4L）领域为例

【摘要】 专利是集技术情报、经济情报、商业情报等于一体的知识载体,其中尤以专利文本中的名称、摘要和主权项为甚,可以凸显专利技术所涉及到的专利知识。伴随着我国在国际科技舞台愈来愈活跃,针对我国海量的专利文本,进行必要的专利知识计量分析,可以有效揭示我国专利技术的知识结构及其发展状况。因此,如何基于我国国家知识产权局专利数据库(the State Intellectual Property Office Database, SIPOD)独特的专利文本结构,探寻适宜的专利知识计量指标体系、方法和应用,就成了本文研究的主要目的。首先,以知识、知识计量学基本理论及相关概念为基础,立足于专利知识的创造性、系统性、网络性等特征,本文提出了专利知识计量的概念,并阐释了“知识元”以及知识元相互联系、相互作用而形成的“知识链”、“知识群”和“知识网络”等各类知识聚集体是专利知识计量的对象。之后,构建了专利知识元、专利知识链、专利知识群和专利知识网络四层次专利知识计量指标体系。其中,在专利知识元角度,采用指标频数、权重、度数和中介中心度从其数量和质量角度进行了衡量：在专利知识链方面,则从频数、链长、影响力和中介中心度角度对知识链的构成情况及作用进行了度量；在专利知识群方面,是以强度、规模和聚集度指标对它的结构和功能进行了比较；在专利知识网络方面,则以规模、平均最短距离和密度指标对其组成和结构进行了分析。其次,针对我国SIPOD的专利文本结构,在最大字符串匹配算法的基础上,结合具体专利知识元所在的名称、摘要和主权项位置,构建了专利知识元抽取的方法,并采用知识元云图的方式对其进行了知识可视化。之后,本文采用关联规则挖掘算法Apriori中的频繁项集抽取了专利知识链,并借鉴复杂网络分析中凝聚子群的识别法,以Lambda集合算法完成了专利知识群的识别与计算。本文提到的专利知识网络是以抽取出来的专利知识元为节点、以知识元之间的余弦相似度为边形成的,之后借鉴复杂网络的思想对其结构进行了分析。最后,以与下一代移动通信网络密切相关的数字信息的传输(H04L)领域中76877项专利文本为研究对象,开展了专利知识元、专利知识链、专利知识群和专利知识网络四个层次的专利知识计量分析。进而从专利知识元、专利知识链、专利知识群等角度展示了“H04L”技术领域的专利知识结构及各层次问的知识联系,之后结合专利知识元、专利知识链和专利知识群的特征,采用相应的计量指标定位了"H04L"领域的核心技术知识、热点技术知识、关键技术知识和核心技术知识群。更多还原

【Abstract】 Patent is a kind of knowledge carrier concentrating on technological information, economic information, business information and other forms of information. Especially, as the title, abstract and claim of the patent text, which could be analyzed to reveal its technical knowledge. With the development of science and technology of China, knowledge-based quantitative analysis on volumes of patent texts is able to demonstrate the knowledge construction and development status of patent technology of our country. Based on the particular patent structure in the SIPOD, the paper aims to find the proper metrics method and indicator to calculate the patent knowledge.First of all, based on the fundamental theory and relevant concept of knowledge and knowmetrics, the paper defines the patent knowmetrics with patent knowledge’s features established in inventiveness, systematicness, and networking. Metric object of patent knowmetrics is explained based on the knowledge unit, because kinds of knowledge units interacting together assemble the stratiform cluster knowledge link, knowledge group and knowledge network. Therefore, a construction of indicator system has been established with a four level structure of patent knowledge unit, knowledge link, knowledge group and knowledge network. In the perspective of patent knowledge unit, the frequency, weight, degree and betweenness are applied to measure its quantity and quality. On the respect of knowledge link, the indicators frequency, length, impact and betweenness are formed to analyze its structure and function. In the field of patent knowledge group, the indicator strength, size and cluster are used to make comparison among a large of groups. Furthermore, the size, distance and density are adopted to describe the feature of patent knowledge network.Secondly, the paper designs different methods to extract different patent knowledge in different levels, based on the unique patent structure in the SIPOD. In the level of knowledge unit, the algorithm of maximum string matching is applied, combining with the position of knowledge unit, and then cloud of knowledge unit is applied to knowledge visualization. In this research, patent knowledge link has been extracted with frequency item set of Apriori mining algorithms in association rules. Taking example by identification methods of complex network analysis, patent knowledge group has been recognized and calculated with Lambda Set from social network analysis. With regard to knowledge network, the analysis method in the complex network is used to show its characteristics.Finally, taking the76,877patents in the H04L field in the SIPOD as the example, closely connecting with the next generation mobile communication and tri-networks integration, knowmetrics analysis is carried out with the indicator system built above. With the results shown in the four levels, the knowledge structure of H04L filed has been revealed from the perspective of knowledge unit, knowledge link, knowledge group and knowledge network. At the same time, the research hotspots in the field has been identified with the indicator frequency and the length in the patent knowledge link, while the indicator cluster in the knowledge group is used to represent the core technological group knowledge. Simultaneously, the core technological knowledge and the key technological knowledge have been positioned.更多还原