Abstract: To obtain knowledge from patent documents to inspire the creative thinking in product innovative design, and to improve product innovation capability, a method of patent text classification oriented to product innovation was proposed. With the application of text mining and natural language understanding, patent text classification based on TRIZ Inventive Principles was studied. Taking United States Patents as data source, by integrating knowledge expression of Invention Principles and text mining, patent characteristics information was extracted. And a uniform patent characteristic representation model was set up. On this basis, visualization software for patent classification was developed with VC++. Finally, relevant patents mined by this software were analyzed. Innovative design of paper machine was conducted and new innovative design schemes were also obtained. And the method was proved to be effective.

Key words: innovative design, patent classification, text mining, inventive principle, theory of invention problem solving, product design

摘要： 为提高专利知识在产品创新设计中的应用价值，以打破思维定势、激发创新灵感，提出了一种面向创新设计的专利文本分类的思路及方法。基于发明问题解决理论（TRIZ），应用文本挖掘和自然语言理解等先进的技术手段，以TRIZ发明原理为分类标准对专利文本进行了自动分类的研究。以美国专利为数据源，将发明原理的知识表示与文本挖掘策略相融合，抽取专利特征信息，建立统一的专利特征表示模型，并使用VC++开发出了相应的软件系统。最后分析了该软件挖掘出的相关专利，对造纸机进行了创新设计，辅助得到了新的原理方案，验证了该方法的有效性。

关键词: 创新设计, 专利分类, 文本挖掘, 发明原理, 发明问题解决理论, 产品设计