首页|融合专利SAO特征的早期专利价值评估方法

融合专利SAO特征的早期专利价值评估方法

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
[研究目的]在知识产权强国战略背景下,存在着专利数量与质量不匹配、技术转化率较低的困境,评估早期专利价值对高价值专利培育工程和专利技术转化具有重要意义.[研究方法]提出一种融合专利SAO网络特征和文本特征的早期专利价值评估方法.首先,从专利摘要和标题中抽取主体-行为-客体(SAO)结构.其次,用社会网络分析法挖掘专利SAO网络的核心组件,得到专利SAO网络中心性特征.然后,将SAO结构序列文本输入Doc2Vec模型,得到专利SAO文本特征.最后,融合专利基础特征和SAO特征,构建基于AutoGluon自动机器学习框架的早期专利价值评估模型.[研究结论]以人工智能产业专利为例进行实证研究,实验结果表明引入专利SAO特征后,模型的准确率、召回率和F1值相比基准模型均提高了 2％～5％,证明了专利SAO特征在识别早期高价值专利中的有效性.
Research on Early Patent Value Evaluation Method Integrating SAO Features
[Research purpose]In the context of the strategy of strengthening China through intellectual property rights,there is a dilem-ma that patent quantity increases but quality decreases,and technology transfer rate is low.Evaluating the value of early patents is of great significance for the cultivation of high-value patents and the transfer of patent technologies.[Research method]We propose an early pa-tent value assessment method that integrates features from the patent SAO(Subject-Action-Object)network and text.Firstly,we extract the SAO structures from patent abstracts and titles.Secondly,we employ social network analysis to identify the core components of the pa-tent SAO network and obtain centrality features.Next,we input the SAO structure sequence text into the Doc2Vec model to derive the pa-tent SAO text features.Finally,we combine the basic patent features with the SAO features to construct an early patent value assessment model based on the AutoGluon automated machine learning framework.[Research conclusion]Using patents in the field of artificial intel-ligence industry as an example,an empirical research is carried out.Experimental results demonstrate that the introduction of patent SAO fea-tures improves the accuracy,recall,and Fl score of the model by 2-5％compared to the baseline model,which confirms the effectiveness of patent SAO features in identifying early-stage high-value patents.

patent valueearly patent value evaluationhigh value patentspatent textsartificial intelligencepatent SAO networkau-tomated machine learningsocial network analysis

周小琴、蔡鸿宇、石进、卢明欣

展开 >

南京大学信息管理学院 南京 210023

南京大学(苏州)高新技术研究院 苏州 215127

专利价值 早期专利价值评估 高价值专利 专利文本 人工智能 专利SAO网络 自动机器学习 社会网络分析

国家社会科学基金项目

21BTQ012

2024

10.3969/j.issn.1002-1965.2024.10.023

情报杂志
陕西省科学技术信息研究所

情报杂志

CSTPCDCSSCICHSSCD北大核心
影响因子：1.502
ISSN：1002-1965
年,卷(期)：2024.43(10)
  • 1