ILRD:一种融合IPC_LDA和文本语义特征的专利推荐模型  

ILRD: A Patent Recommendation Model Fusing IPC_LDA and Textual Semantic Features

作　　者：邓娜[1] 林涵辉 陈旭 刘树栋 Deng Na;Lin Hanhui;Chen Xu;Liu Shudong(School of Computer Science,Hubei University of Technology,Wuhan 430068;School of Information Engineering,Zhongnan University of Economics and Law,Wuhan 430073)

机构地区：[1]湖北工业大学计算机学院,武汉430068 [2]中南财经政法大学信息工程学院,武汉430073

出　　处：《情报杂志》2025年第3期170-179,F0003,共11页Journal of Intelligence

基　　金：国家社会科学基金一般项目“人工智能时代新闻推荐系统中的‘信息茧房’问题形成机理及解决策略研究”(编号:21BXW076)研究成果。

摘　　要：[研究目的]随着全球专利数量的持续增长,现有的专利推荐模型难以精准地理解专利文本中复杂的语义信息,忽视了专利的领域和主题信息,从而影响专利的准确推荐。因此,实现一个能够综合考虑多种因素的高效专利推荐模型具有重要意义。[研究方法]提出了一种ILRD专利推荐模型,通过IPC语义结构化映射与LDA主题建模来表征专利的技术领域和核心主题。构建RoBERTa与DPCNN模型深入解析专利文本的复杂语义,以实现多因素融合的专利推荐。[研究结果/结论]实证研究显示,ILRD模型在半导体领域的专利推荐上,其准确率和F1得分显著超越常规模型,推荐准确率提高至86%。验证了本文方法的有效性,也为跨领域专利推荐提供了新思路。[Research purpose]With the continuous growth in the number of global patents,existing patent recommendation models struggle to accurately interpret the complex semantic information within patent texts,often overlooking the domain-specific and thematic information that is crucial for precise recommendations.Developing an effective patent recommendation model that comprehensively considers various factors is therefore of significant importance.[Research method]This study introduces the ILRD patent recommendation model,which uses IPC semantic structured mapping and LDA topic modeling to delineate the technical domains and core themes of patents.The RoBERTa and DPCNN models are employed to deeply analyze the complex semantics of patent texts,facilitating a multi-faceted fusion approach to patent recommendations.[Research result/conclusion]Empirical results demonstrate that the ILRD model substantially surpasses conventional models in accuracy and F1 score for patent recommendations in the semiconductor field,enhancing recommendation accuracy to 86%.This validates the effectiveness of the proposed method and offers new perspectives for cross-domain patent recommendations.

关 键 词：专利推荐 IPC LDA主题建模 预训练模型 RoBERTa DPCNN 半导体 

分 类 号：G306[文化科学] TP18[自动化与计算机技术—控制理论与控制工程]