基于LDA和深度模型的专利主题演化
Patent Topic Evolution Based on LDA and Deep Models
关菁华 1毛子凡 2王婷 1谭梦琪1
作者信息
- 1. 大连外国语大学软件学院,辽宁 大连 116044
- 2. 大连东软信息学院图书馆,辽宁 大连 116023
- 折叠
摘要
将基于深度模型的命名实体识别方法与基于LDA模型的主题挖掘方法相结合进行专利技术主题演化分析,满足用户和企业从众多专利中快速找到主要技术方法和热门研究领域的需求.以环氧树脂专利为例,进行基于深度学习的相关化合物及制备技术实体识别,采用半自动化方式构建领域词典,再使用改进的LDA模型进行主题建模,从主题强度及内容两个角度分析演化过程.半自动化领域词典的构建,能使主题挖掘结果更加准确,改进的LDA模型有助于主题内容的解释,为主题演化提供新的思路与方法.
Abstract
The named entity recognition method based on the depth model is combined with the topic mining method based on the LDA model to analyze the evolution of patented technical subjects,so as to meet the needs of users and enterprises to quickly find the main technical methods and hot research fields from many patents.Taking the epoxy resin patent as an example,the entity identification of related compounds and preparation technologies based on deep learning is carried out,and the domain dictionary is constructed in a semi-automatic way.The improved LDA model is used to model,and the evolution process is analyzed from the two directions of subject strength and content.The construction of semi-automatic domain dictionary can make the topic mining results more accurate.The improved LDA model is helpful to the interpretation of topic content and provides new ideas and methods for topic evolution.
关键词
LDA/深度模型/主题演化/专利Key words
LDA/deep model/topic evolution/patent引用本文复制引用
基金项目
大连外国语大学科研项目(2021XJYB18)
辽宁省社会科学规划基金(L17BTQ005)
出版年
2024
NETL
NSTL
万方数据