摘要
冶金行业为集成电路、人工智能、航空航天等重要领域提供必不可少的金属原材料。冶金国家标准是重要的基础性战略资源,在冶金产业的高质量发展中发挥技术性支撑作用。国家标准文本中蕴含大量的关键技术性指标,人工逐一识别并抽取的模式在大数据时代已无法满足数字化转型的需求。本研究采用深度学习算法,对冶金领域的国家标准文本分别开展RNN、GRU和BiLSTM模型的对比实验,根据模型性能择优选取最优模型。研究结果显示BiLSTM模型在冶金领域国家标准指标识别的表现上最好,由此采用BiLSTM为该领域标准指标识别的深度学习模型。
Metallurgy plays a fundamental role in providing indispensable metal raw materials for important industries such as integrated circuit, artificial intelligence and aerospace. Metallurgical national standards are basic strategic resources of great importance, supporting the high-quality development of metallurgy. The contents of national standards contain a large number of critical technical indicators. Manually identifying and extracting indicators fail to meet the requirement of digital transformation after the advent of the Big Data Era. The deep learning models are used to conduct 3 experiments based on RNN, GRU, and BiLSTM model on metallurgical national standards to find an optimal solution.The results suggested that BiLSTM model performed best in the identification of indicators in metallurgical national standards, thus making BiLSTM the solution to the identification of standard indicator in this field.
作者
夏磊
方思怡
解凌
蔡焱
顾晓虹
XIA Lei;FANG Si-yi;XIE Ling;CAI Yan;GU Xiao-hong(Shanghai Institute of Quality and Standardization)
出处
《中国标准化》
2023年第3期87-93,共7页
China Standardization
基金
上海市市场监督管理局科技项目“标准指标智能抽取和比对技术在政府监管和‘企标领跑者’制度实施中的研究与应用”(项目编号:2021-47)资助。
关键词
冶金
国家标准
标准指标识别
深度学习
BiLSTM
metallurgy
national standard
identification of national standard indicator
deep learning
BiLSTM
