Abstract: Implicit Technical Characteristics (TC) are technical objectives hidden from design requirements and expressed unstructured.Identifying and ranking implicit TCs is the key to responding quickly to technological innovation requirements and ensuring the design value of the product.However,the identification of implicit TCs relies on design experience,which affects the objectivity and extensiveness of TC identification.Besides,the traditional TCs ranking process ignores the innovative value of technological development.To this end,a product implicit technical characteristics identification and ranking based on patent text was proposed.Based on the patent text,the Latent Dirichlet Allocation (LDA) topic model and word vector were utilized to mine the implicit TCs associated with user requirements.The frequency-time matrix of TC words was constructed,and the Markov chain was used to reason out the core TC words that were technologically superior.Then the potential value of TCs was solved based on novelty and technology maturity and ranked comprehensively.Finally,a steam oven design was used to verify the validity of the proposed model,and a prototype system to assist in the identification and ranking of implicit TCs was developed to improve design efficiency.

Key words: technical characteristics, patent text, latent Dirichlet allocation topic model, conceptual design, Markov chain

摘要： 隐性技术特征(TC)是隐藏于设计需求以非结构化表达的技术目标,对隐性TC的识别结果是排序是快速响应技术创新需求并确保产品设计价值的关键。然而,隐性TC的识别依赖于设计经验,影响TC识别结果的客观性和广泛性。此外,传统TC排序过程忽略了技术发展的潜在创新价值。为此,提出基于专利文本的产品隐性TC识别与排序模型。首先,基于专利文本,利用隐含狄利克雷分布(LDA)主题模型和词向量挖掘关联设计需求的TC词。其次,构建TC词的频次时间矩阵,利用马尔科夫链推理出核心隐性TC词。基于新颖度和技术成熟度求解TC的潜在价值并综合排序。最后,以蒸汽烤箱设计为实例验证该模型的有效性,开发辅助隐性TC识别与排序的原型系统来提升设计效率。

关键词: 技术特征, 专利文本, LDA主题模型, 概念设计, 马尔科夫链