现代情报2025，Vol.45Issue(11)：5-17,13.DOI:10.3969/j.issn.1008-0821.2025.11.002
多特征融合视角下基于提示学习的高价值"卡脖子"专利识别
High-Value'Bottleneck'Patent Identification From a Multi-Feature Fusion Perspective Based on Prompt Learning
摘要
Abstract
[Purpose/Significance]This study aims to effectively identify high-value"bottleneck"technology patents by combining adaptive prompt learning with quantitative indicators and semantic content in patents.[Method/Process]A novel identification model for high-value"bottleneck"technology patents was proposed,leveraging carefully designed prompt learning templates to integrate quantitative indicators with patent text,thereby enhancing the model's understand-ing of both quantitative and qualitative data.Additionally,adaptive prompt learning was used to improve the model's abi-lity to interpret complex data.[Result/Conclusion]A patent value identification dataset is constructed,and an empirical study is conducted based on it.Experimental results demonstrate that the use of adaptive prompt learning and the integra-tion of multiple features significantly enhance the model's performance in identifying high-value"bottleneck"technology patents.This research holds significant value in addressing"bottleneck"technology challenges and advancing high-quality development in intellectual property.关键词
提示学习/专利价值/价值评估/专利识别/"卡脖子"技术Key words
prompt learning/patent value/value evaluation/patent identification/"bottleneck"technology分类
社会科学引用本文复制引用
张殿元,余传明..多特征融合视角下基于提示学习的高价值"卡脖子"专利识别[J].现代情报,2025,45(11):5-17,13.基金项目
国家自然科学基金面上项目"基于知识增强的科技文献创新识别与评价模型研究"(项目编号:72374219) （项目编号:72374219）
国家自然科学基金面上项目"面向跨语言观点摘要的领域知识表示与融合模型研究"(项目编号:71974202) （项目编号:71974202）
教育部人文社会科学研究一般项目"多语言情境下基于深度表示与对齐的观点摘要研究"(项目编号:19YJC870029). （项目编号:19YJC870029）
