情报杂志2019，Vol.38Issue(10)：168-175,8.DOI:10.3969/j.issn.1002-1965.2019.10.024
基于深度迁移学习的技术术语识别——以数控系统领域为例
Deep Transfer Learning for Technical Term Extraction—A Case Study in Computer Numerical Control System
摘要
Abstract
Emerging term identification is a significant task in emerging technology forecasting. As patent data constitutes up-to-date and reliable source of technological intelligence,it has been widely applied to emerging technology foresight. How-ever,though this approach to technological forecasting can be easily used,lack of term labels is the critical issue on account of the difficul-ties in term data mining and extraction. Nevertheless,there is few research concerning the use of named entity recognition in term extrac-tion,especially in the field of patent data. [Method/Process]In this paper,we introduce the idea of deep transfer learning and use the pub-lic-domain data,where a source task with plentiful annotations; we attempt to achieve cross-domain transfer,realize the extraction of tech- nical terms and filter high-frequency non-term words effectively by applying Bi-LSTM (Bi-directional Long Short Time Memory) Mod-el. Furthermore,we produce classification of identified technical terms by clustering. [Result/Conclusion]In this paper,through the experi-ment with patent data in the field of CNC (Computer Numerical Control) system,we find that the model can transfer the existing knowl-edge from source-domain public data to target-domain science data. It suggests that the model can solve the issue of lack of term labels in patent data and improve the correlation of extracted term between domains. On this basis,the classification of technical terms can provide data support for CNC filed to predict the trends of technology development.关键词
新兴技术预见/命名实体识别/深度迁移学习/数控系统/专利分析Key words
emerging technology forecasting/named entity recognition (NER)/deep transfer learning/computer numerical control(CNC)/patent analysis分类
社会科学引用本文复制引用
刘宇飞,尹力,张凯,杨建中,郑文江..基于深度迁移学习的技术术语识别——以数控系统领域为例[J].情报杂志,2019,38(10):168-175,8.基金项目
国家自然科学基金项目"支持技术预见的多源异构大数据融合与时序文本预测方法研究" (编号:91646102) （编号:91646102）
国家自然科学基金项目"面向2035的高端装备领域技术路线图总体框架及重点子领域研究"(编号:L1824039) （编号:L1824039）
国家自然科学基金项目"衍化升级情境下2035智能制造领域技术路线图应用研究:基于融合与派生路径" (编号:L172400022) （编号:L172400022）
教育部人文社会科学项目(编号:16JDGC011) （编号:16JDGC011）
中国工程科技知识中心建设项目"工程科技战略咨询智能支持系统建设"(编号:CKCEST-2019-2-13)研究成果之一. （编号:CKCEST-2019-2-13）
