面向TRIZ理论的深度知识获取及应用研究

发布时间：2018-12-31 21:33

【摘要】：传统的产品设计过程对概念设计阶段创新力不足，设计人员往往由于自身单一的专业知识很难产生真正创新的产品。TRIZ理论虽然能够指导设计人员创新，但在实际应用中不易被完全掌握。专利是产品创新的主要知识资源，但是传统专利库中包含了不计其数的专利文件，且是以学科作为分类基础的，难以被设计人员查找和利用。因此按照TRIZ理论相关知识提取、重组以及分析，以及对已有知识库的扩充和更新有利于TRIZ理论的实际应用和自身完善，从创新理论的角度上辅助人们掌握创新设计的普遍规律，在类比以往成功案例的基础上激发设计人员的发散思维是非常有意义的。提出了基于TRIZ理论的深度知识获取模型。该模型是以TRIZ理论为基础，采用数据挖掘技术为手段，利用中文专利文献资源来获取深度知识的。专利的深度知识获取研究有助于专利知识跨学科应用和知识发现与重用研究，以及TRIZ理论由理论的高度变成被普通设计者接受的一般的理解并实际应用过程的探索。论文按照基于TRIZ理论中文专利深度知识获取模型的各子模块的实现方法和关键技术进行论述，包括以下几个部分：专利文本抽取模块、文本分类器模块和深度知识挖掘模块。首先，专利文本抽取模块介绍了如何在国家知识产权局这样的网页中获取所需要的专利库信息，并将之保存到数据库当中。专利文本抽取是整个知识挖掘过程的前提，如果不能正确抽取所需要的专利摘要和基本信息，就无法构建专利库，进行下一步的分析，因此，本模块的研究对于后续的研究至关重要。其次，文本分类器模块主要实现了对专利文本的分类过程。将从人工分类和计算机分类两个方面阐述专利分类的原理及过程。人工辅助专利分类是建立在人工仔细阅读专利说明书的基础上进行的，要求分类人员掌握一定的TRIZ知识和相关的领域知识。而计算机分类的对象都是专利文档的摘要部分，摘要既能基本上代替专利全文的基本内容，而且对于计算机计算的难度大大简化，对于分类器在测试阶段是非常简便和实用的。本模块的专利分类主要是指运用发明原理为分类背景对专利进行分类和分析的。再次，深度知识挖掘模块主要利用了分类的结果进行深度知识提取。挖掘过程是在阅读专利说明书的基础上，按照深度知识模板向导进行分析的，最后将分析的结果存入实例库中。最后，构建了该深度知识获取模型的软件系统——DKMining。该软件系统实现了各模块的功能，，并且能对专利库和实例库中的信息实现检索、删除、修改、更新功能。论文结合具体专利实例验证上述理论研究，因此软件系统具有一定的可行性。
[Abstract]:The traditional product design process has insufficient innovation ability to the conceptual design stage, the designer is often difficult to produce the truly innovative product because of their own single professional knowledge. Although the TRIZ theory can guide the designer's innovation, But in the practical application is not easy to be fully grasped. Patent is the main knowledge resource of product innovation, but the traditional patent library contains countless patent files, and is based on disciplines, so it is difficult to be searched and utilized by designers. Therefore, according to the relevant knowledge extraction, reorganization and analysis of TRIZ theory, as well as the expansion and updating of the existing knowledge base, it is beneficial to the practical application and self-improvement of the TRIZ theory. From the perspective of innovation theory, it helps people to grasp the general law of innovative design. It is very meaningful to stimulate the divergent thinking of designers on the basis of analogizing past successful cases. A deep knowledge acquisition model based on TRIZ theory is proposed. The model is based on TRIZ theory and uses data mining technology to acquire deep knowledge by using Chinese patent literature resources. The research of patent in-depth knowledge acquisition is helpful to the interdisciplinary application of patent knowledge and the research of knowledge discovery and reuse, as well as the exploration of TRIZ theory from a theoretical height to a general understanding and practical application accepted by ordinary designers. According to the implementation method and key technology of the Chinese patent depth knowledge acquisition model based on TRIZ theory, this paper discusses the following parts: patent text extraction module, text classifier module and depth knowledge mining module. First, the patent text extraction module introduces how to obtain the required patent library information in a web page such as the State intellectual property Office, and store it in the database. Patent text extraction is the premise of the whole process of knowledge mining. If the patent abstract and basic information can not be extracted correctly, the patent database can not be constructed and the next step can be analyzed. The research of this module is very important for further research. Secondly, the text classifier module mainly realizes the patent text classification process. This paper expounds the principle and process of patent classification from two aspects: manual classification and computer classification. Artificial assistant patent classification is based on manual careful reading of patent specification, which requires classifier to master certain TRIZ knowledge and related domain knowledge. The objects of computer classification are abstracts of patent documents, which can not only replace the basic contents of patent full text, but also simplify the difficulty of computer calculation greatly, and it is very simple and practical for classifiers in the testing stage. The patent classification of this module mainly refers to the classification and analysis of patents using the invention principle as the classification background. Thirdly, the deep knowledge mining module mainly uses the classification results to extract the depth knowledge. On the basis of reading patent specification, mining process is analyzed according to the deep knowledge template guide. Finally, the results of the analysis are stored in the case library. Finally, a software system named DKMining. is constructed for the deep knowledge acquisition model. The software system realizes the function of each module, and can retrieve, delete, modify and update the information in patent library and instance library. The paper verifies the above theory research with the concrete patent example, therefore the software system has certain feasibility.
【学位授予单位】：陕西科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.1;TB472

【参考文献】