|
|
|
| Analysis of Patent Technology Topic Evolution Based on Product Life Cycle |
| Ma Jianhong1, Wang Chenxi1, Yan Lin2, Yao Shuang1 |
1.School of Artificial Intelligence, Hebei University of Technology, Tianjin 300401
2.Tianjin Gongchuang Technology Development Co., Ltd., Tianjin 300000 |
|
|
|
|
Abstract The evolution of the patent technology subject contains the development context of product technology. Grasping it accurately is very important for product technology researchers. In view of the shortcomings of the existing patent technology subject evolution analysis, a patent technology subject evolution analysis method based on product life cycle is proposed. First, aiming at the problem of the existing time feature partitioning method leading to an inundation of topic information in the early stage of product development, a product life cycle partitioning method is proposed that uses the patent growth law and local topic semantic similarity to divide the product life cycle. Second, aiming at the problems of poor applicability of the existing subject mining methods to the patent literature and the lack of obvious technical characteristics of the subject, a Multiple Weighted Latent Dirichlet Allocation (MW-LDA) patent subject extraction method is proposed. MW-LDA is used to improve the term weights from multiple angles and construct compound weights to optimize the process by which an LDA model generates feature words, making it more applicable to patent texts. Then, a method for the evolution analysis of the product patent technology topic is proposed, and the evolution analysis of the product patent technology topic in different stages of the product life cycle is realized using the strength of topic correlation between phases. Finally, the experimental results and comparative experiments on a corpus of patents of electric vehicle power units show that the method can effectively identify the technical topics of patents and analyze the development trend of products, which can provide support for scientific research and science and technology policymaking.
|
|
Received: 14 May 2021
|
|
|
|
1 陈亮. 面向专利分析的Patent Classification LDA模型[J]. 情报学报, 2016, 35(8): 864-874.
2 秦旭, 杨文忠, 王雪颖, 等. 基于共现关系的多源主题融合模型[J]. 计算机工程与应用, 2020, 56(10): 157-162.
3 刘啸剑, 谢飞, 吴信东. 基于图和LDA主题模型的关键词抽取算法[J]. 情报学报, 2016, 35(6): 664-672.
4 Othman R, Noordin M F, Gusmita R H, et al. SAO extraction on patent discovery system development for Islamic finance and banking[C]// Proceedings of the 2016 6th International Conference on Information and Communication Technology for the Muslim World. IEEE, 2016: 59-63.
5 Blei D M. Probabilistic topic models[J]. Communications of the ACM, 2012, 55(4): 77-84.
6 Wang A, Zhang J J. Topic discovery method based on topic model combined with hierarchical clustering[C]// Proceedings of the 2020 IEEE 5th Information Technology and Mechatronics Engineering Conference. IEEE, 2020: 814-818.
7 李海林, 邬先利. 基于时间序列聚类的主题发现与演化分析研究[J]. 情报学报, 2019, 38(10): 1041-1050.
8 程磊, 高茂庭. 结合时间加权和LDA聚类的混合推荐算法[J]. 计算机工程与应用, 2019, 55(11): 160-166.
9 彭云, 万常选, 江腾蛟, 等. 基于语义约束LDA的商品特征和情感词提取[J]. 软件学报, 2017, 28(3): 676-693.
10 陈伟, 林超然, 李金秋, 等. 基于LDA-HMM的专利技术主题演化趋势分析——以船用柴油机技术为例[J]. 情报学报, 2018, 37(7): 732-741.
11 Guo H C, Liang Q L, Li Z Q. An improved AD-LDA topic model based on weighted Gibbs sampling[C]// Proceedings of the 2016 IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference. IEEE, 2016: 1978-1982.
12 李湘东, 巴志超, 黄莉. 基于加权隐含狄利克雷分配模型的新闻话题挖掘方法[J]. 计算机应用, 2014, 34(5): 1354-1359.
13 吴菲菲, 陈肖微, 黄鲁成, 等. 基于语义相似度的技术多主题演化路径识别方法研究[J]. 情报杂志, 2018, 37(5): 91-96.
14 Cong H, Tong L H. Grouping of TRIZ Inventive Principles to facilitate automatic patent classification[J]. Expert Systems with Applications, 2008, 34(1): 788-795.
15 乜丽丽. 基于专利分析的技术成熟度预测方法研究与实现[D]. 天津: 河北工业大学, 2011.
16 Prabhudesai K S, Mainsah B O, Collins L M, et al. Augmented latent Dirichlet allocation (LDA) topic model with Gaussian mixture topics[C]// Proceedings of the 2018 IEEE International Conference on Acoustics, Speech and Signal Processing.. IEEE, 2018: 2451-2455. |
|
|
|
