利用技术功效语义关联构建技术实现路径  

Technology Implementation Path Based on Technology Effect Semantic Correlation

作　　者：张金柱[1] 于文倩 李溢峰 ZHANG Jinzhu;YU Wenqian;LI Yifeng

机构地区：[1]南京理工大学经济管理学院

出　　处：《图书馆论坛》2021年第3期31-41,共11页Library Tribune

基　　金：国家自然科学基金项目“基于表示学习的专利信息语义融合与深度挖掘研究”(项目编号:71974095);江苏省社会科学基金项目“基于社团结构动态演化的主题突变监测与形成机制研究”(项目编号:17TQC003);国家自然科学基金项目“基于被引科学知识突变的突破性创新动态识别及其形成机理研究”(项目编号:71503125)研究成果。

摘　　要：文章明晰技术功效间的多种语义联系,设计技术实现路径的自动化构建方法,实现其即时更新和可视化。结合专利数据特点,基于规则从专利标题中抽取技术词,利用BiLSTM-CRF深度学习模型从专利摘要中抽取专利功效短语,并设计规则从功效短语中自动识别出功效词以及表示技术功效间语义联系的关系词,构建“技术词-关系词-功效词”结构的技术功效语义关联,通过计算实体间语义相似度实现技术词对齐和功效词对齐,优化技术功效关联,依此构建技术实现路径,并以知识网络的形式对其进行可视化。在5G技术领域的实证结果表明,该方法能有效揭示技术功效间的多种语义联系和自动构建技术实现路径,并实现路径的即时更新和清晰展示。This paper clarifies the multiple semantic relationship between technology and effect,devises the automatic construction method of technology implementation path,and realizes its live-update and visualization.Firstly,according to the characteristics of patent data,it extracts technology topic terms based on certain rules and patent effect phrases using BiLSTM-CRF deep learning model respectively.Secondly,it designs rules to automatically identify effect words from effect phrases and relation words representing semantic relations between technology and effect,and constructs a technology effect correlation representation of the"technology wordsrelation words-effect words"structure.Finally,it realizes the alignment of technology words and effect words by calculating semantic similarity between entities,and optimizes and visualizes the technology implementation path in the form of knowledge network.The empirical results in 5G technology field show that the method can effectively reveal various semantic relationships between technology and effect,and has a high degree of automation,which is convenient for instant updates and clear display of technology effect correlation.

关 键 词：技术功效 实现路径 信息抽取 专利分析 

分 类 号：G306[文化科学]