机械产品专利知识的提取和应用

Extraction and Application of Mechanical Product Patent Knowledge

随着专利数量的快速增长,单纯依靠人工进行专利查阅,很难及时、快速获取专利中的创新知识资源。对此,引入深度学习方法,研究机械产品专利知识的计算机自动抽取,进而实现机械产品的创新设计。在研究中,将专利知识抽取分解为对专利中功效、原理、结构三类实体的识别,以及各类实体之间关系的抽取,由此构建专利知识结构模型。基于BERT语言预训练模型完成实体的识别和实体关系的抽取,解决了传统专利分析方法抽取知识片面且精度不高、效率较低的问题。与此同时,设计了对机械产品创新设计起辅助作用的专利知识推送系统。通过电动牙刷实例验证了研究的有效性。

With the rapid increase in the number of patent,it is difficult to obtain innovative knowledge resource in patent in a timely and rapid manner by relying solely on manual patent review.In this regard,the deep learning method was introduced to explore the automatic extraction of patent knowledge on mechanical products by computer,and realize the innovative design of mechanical product.In the research,the patent knowledge extraction is decomposed into the identification of three types of entities in the patent such as efficacy,principle,and structure,as well as the extraction of the relationship between various entities,thereby constructing a patent knowledge structure model.Based on the BERT language pre training model,the identification of entity and extraction of entity relationship are completed,which solves the problem of low precision and low efficiency extraction of one sided knowledge by traditional analytical approach for patent.At the same time,a patent knowledge push system was designed to assist in the innovative design of mechanical product.The effectiveness of the research was verified by an example of an electric toothbrush.