基于自然语言处理的CNAS认可准则自动对标系统

doi:10.16180/j.cnki.issn1007-7820.2023.05.005

摘要/Abstract

摘要：

在CNAS评审过程中,人工对标不符合项与其依据条款存在着耗时耗力、对标不准确等缺点。文中针对以上问题,提出一种基于自然语言处理的多级模型自动对标方法。通过对不符合项描述的语言特点进行研究,利用注意力机制的Bi-LSTM网络对不符合项进行分类。在该分类下使用基于语料扩充和迁移学习的SimCSE网络模型计算相似的不符合项,并提取对应依据条款,有效解决了对标不准确等问题。模拟实验测试表明,所提方法的对标准确率可达74.4%,语义匹配计算时间相比DSSM模型有大幅提升,内存消耗和最高匹配速度也有明显改善。

关键词: 深度学习, 自然语言处理, 语义匹配, 自动对标, 多标签分类, 多模型融合, 注意力机制, 语义计算

Abstract:

During the CNAS review process, manual benchmarking of non-conformance items and its compliance clauses has the disadvantages of time-consuming, labor-intensive, and inaccurate benchmarking. In view of the above problems, this study proposes a multi-level model automatic benchmarking method based on natural language processing. By studying the linguistic characteristics of the description of non-conforming items, the Bi-LSTM network of the attention mechanism is used to classify non-conforming items. Under this classification, the SimCSE network model based on corpus expansion and transfer learning is used to calculate similar non-conforming items and extract the corresponding basis clauses, which effectively solves the problems such as inaccurate benchmarking.Through simulation experiments, the benchmarking accuracy rate of the proposed method can reach 74.4%, and the semantic matching calculation time is greatly improved when compared with the DSSM model, and the memory consumption and the highest matching speed are also been significantly improved.

Key words: deep learning, natural language processing, semantic matching, benchmarking system, multi label classification, multi model fusion, attention mechanism, semantic computing

中图分类号:

TP391

刘玉威,曹民,冯浩甲. 基于自然语言处理的CNAS认可准则自动对标系统[J]. 电子科技, 2023, 36(5): 28-33.

LIU Yuwei,CAO Min,FENG Haojia. CNAS Recognition Criteria Automatic Benchmarking System Based on Natural Language Processing[J]. Electronic Science and Technology, 2023, 36(5): 28-33.

图/表 8

图1

图2

图3

表1

表2

图4

图5

表3

参考文献 19

[1]	中国合格评定国家认可委员会.ISO/IEC17025:2018. 检测和校准实验室能力认可准则[S]. 北京: 中国标准出版社,2019.
	China National Accreditation Commission for Conformity Assessment.ISO/IEC17025:2018. Accreditation criteria for the competence of testing and calibration laboratories[S]. Beijing: Standards Press of China,2019.
[2]	中国合格评定国家认可委员会.ISO/IEC17025:2020. 认可标识使用和认可状态声明规则[S]. 北京: 中国标准出版社,2020.
	China National Accreditation Commission for Conformity Assessment.ISO/IEC17025:2020. Rules for use of approval marks and declaration of approval status[S]. Beijing: Standards Press of China,2020.
[3]	中国合格评定国家认可委员会.ISO/IEC17025:2018. 公正性和保密规则[S]. 北京: 中国标准出版社,2018.
	China National Accreditation Commission for Conformity Assessment.ISO/IEC17025:2018. Fairness and confidentiality rules[S]. Beijing: Standards Press of China,2018.
[4]	中国合格评定国家认可委员会.ISO/IEC17025: 2019.申诉、投诉和争议处理规则[S]. 北京: 中国标准出版社,2019.
	China National Accreditation Commission for Conformity Assessment.ISO/IEC17025:2019. Rules for handling complaints, complaints and disputes[S]. Beijing: Standards Press of China,2019.
[5]	中国合格评定国家认可委员会.ISO/IEC 17025:2019. 实验室认可规则[S]. 北京: 中国标准出版社,2020.
	China National Accreditation Commission for Conformity Assessment.ISO/IEC 17025:2019. Laboratory accreditation rules[S]. Beijing: Standards Press of China,2020.
[6]	郝超, 裘杭萍, 孙毅, 等. 多标签文本分类研究进展[J]. 计算机工程与应用, 2021, 57(10):48-56. doi: 10.3778/j.issn.1002-8331.2101-0096
	Hao Chao, Qiu Hangping, Sun Yi, et al. Research progress of multi-label text classification[J]. Computer Engineering and Applications, 2021, 57(10):48-56. doi: 10.3778/j.issn.1002-8331.2101-0096
[7]	武永亮, 赵书良, 李长镜, 等. 基于TF-IDF和余弦相似度的文本分类方法[J]. 中文信息学报, 2017, 31(5):138-145.
	Wu Yongliang, Zhao Shuliang, Li Changjing, et al. Text classification method based on TF-IDF and cosine similarity[J]. Journal of Chinese Information Technology, 2017, 31(5):138-145.
[8]	Li X, Liu N, Yao C, et al. Text similarity measurement with semantic analysis[J]. International Journal of Innovative Computing Information & Control, 2017, 13(5):1693-1708.
[9]	卢佳伟, 陈玮, 尹钟. 融合TextRank算法的中文短文本相似度计算[J]. 电子科技, 2020, 33(10):51-56.
	Lu Jiawei, Chen Wei, Yin Zhong. Chinese short text similarity calculation based on textrank algorithm[J]. Electronic Science and Technology, 2020, 33(10):51-56.
[10]	Chen G L, Shi X L, Chen M K. Text similarity semantic calculation based on deep reinforcement learning[J]. International Journal of Security and Networks, 2020, 15(1):22-29.
[11]	胡艳霞, 王成, 李弼程, 等. 基于多头注意力机制Tree-LSTM的句子语义相似度计算[J]. 中文信息学报, 2020, 34(3):23-33.
	Hu Yanxia, Wang Cheng, Li Bicheng, et al. Sentence semantic similarity computation based on Tree-LSTM with multi-head attention[J]. Journal of Chinese Information Technology, 2020, 34(3):23-33.
[12]	赵琪, 杜彦辉, 芦天亮, 等. 基于Capsule-BiGRU的文本相似度分析算法[J]. 计算机工程与应用, 2021, 57(15):171-177. doi: 10.3778/j.issn.1002-8331.2004-0253
	Zhao Qi, Du Yanhui, Lu Tianliang, et al. Algorithm of text similarity analysis based on Capsule-BiGRU[J]. Computer Engineering and Applications, 2021, 57(15):171-177. doi: 10.3778/j.issn.1002-8331.2004-0253
[13]	杨泉. 基于遗传算法的词语语义相似度计算研究[J]. 计算机技术与发展, 2021, 31(2):8-13.
	Yang Quan. Research on word semantic similarity calculation based on genetic algorithm[J]. Computer Technology and Development, 2021, 31(2):8-13.
[14]	于碧辉, 王加存. 孪生网络中文语义匹配方法的研究[J]. 小型微型计算机系统, 2021, 42(2):231-234.
	Yu Bihui, Wang Jiacun. Research on Chinese semantic matching in siamese network[J]. Journal of Chinese Computer Systems, 2021, 42(2):231-234.
[15]	杨德志, 柯显信, 余其超, 等. 基于RCNN的问题相似度计算方法[J]. 计算机工程与科学, 2021, 43(6):1076-1080.
	Yang Dezhi, Ke Xianxin, Yu Qichao, et al. Problem similarity calculation method based on RCNN[J]. Computer Engineering and Science, 2021, 43(6):1076-1080.
[16]	赵小虎, 赵成龙. 基于多特征语义匹配的知识库问答系统[J]. 计算机应用, 2020, 40(7):1873-1878. doi: 10.11772/j.issn.1001-9081.2019111895
	Zhao Xiaohu, Zhao Chenglong. Knowledge base question answering system based on multi-feature semantic matching[J]. Journal of Computer Applications, 2020, 40(7):1873-1878. doi: 10.11772/j.issn.1001-9081.2019111895
[17]	牛奉高, 高旭霞. 基于加权网络改进的中文短文本相似性度量模型[J]. 情报学报, 2021, 40(3):278-285.
	Niu Fenggao, Gao Xuxia. An improved similarity measurement model for Chinese short texts based on weighted network[J]. Journal of the China Society for Scientific and Technical Information, 2021, 40(3):278-285.
[18]	周练. Word2Vec的工作原理及应用探究[J]. 科技情报开发与经济, 2015, 25(2):145-148.
	Zhou Lian. Exploration of the working principle and application of Word2Vec[J]. Science and Technology Information Development and Economy, 2015, 25(2):145-148.
[19]	朱张莉, 饶元, 吴渊, 等. 注意力机制在深度学习中的研究进展[J]. 中文信息学报, 2019, 33(6):1-11.
	Zhu Zhangli, Rao Yuan, Wu Yuan, et al. Research progress of attention mechanism in deep learning[J]. Journal of Chinese Information Technology, 2019, 33(6):1-11.

标签	不符合描述	不符合条款号	不符合项类别
1	实验室提供不出近2年参加的电气领域/结构判定(电气间隙和爬电距离)子领域能力验证记录。	CNAS-CL01∶2018/7.7.3	不符合认可规则类文件要求
8	实验室未能提供从事电气领域检测人员的安全培训记录。	CNAS-CL11∶2015 5.2.2	人员
12	实验室提供不出设备编号为AL295的直流充电桩模拟分析仪(EV7002)的校准证书。	CNAS-CL01∶2018/6.4.6	溯源性

Sentence1	Sentence2	Label
明天上海天气怎么样	上海明天天气怎么样。	1
我手机丢了,我想换个手机	我想买个新手机,求推荐	1
西宁在哪个省	西宁是哪个市	0

模型	方法	准确率 /%	训练耗时 /min	检索耗时 /s	内存消耗 /G
DSSM	深度语义匹配	73.3	272	22.0	2.7
SVM	支持向量机	60.2	31	5.0	1.2
SimCSE	对比学习	75.5	69	18.0	13.6
本方案模型	双模型融合	74.4	45	3.4	2.5