基于多任务学习的IT运维服务需求语义解析  

Semantic parsing of IT operation and maintenance service requirements based on multi-task learning

作　　者：许明阳 刘振元[1,2] 王承涛 XU Mingyang;LIU Zhenyuan;WANG Chengtao(School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China;Key Laboratory of Education Ministry for Image Processing and Intelligent Control,Wuhan 430074,China;Wuhan Windoor Information Technology Company Co.,Ltd.,Wuhan 430040,China)

机构地区：[1]华中科技大学人工智能与自动化学院,湖北武汉430074 [2]图像信息处理与智能控制教育部重点实验室,湖北武汉430074 [3]武汉问道信息技术有限公司,湖北武汉430040

出　　处：《计算机集成制造系统》2024年第2期673-683,共11页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(72071087);武汉市第四批“黄鹤英才计划”入选人才资助项目。

摘　　要：IT运维服务的自动化水平影响着企业的运营效率,为实现基于无人坐席的智能服务台,提出一种IT运维服务需求语义解析方法,包括意图识别和命名实体识别两个任务。在Multi-BERT-BiLSTM-CRF(MBBC)基准模型之上,通过先验知识和外部资源将词性和实体词典特征融入编码层,增强模型对词法信息和领域知识的学习。对MBBC模型的参数共享方式进行改进,提出增强的MBBC模型模型,增强两个任务之间的信息共享能力。实验表明,与MBBC模型相比,融合词性与实体词典特征并采用增强的MBBC模型可以进一步提升两类任务的识别性能。The automation level of IT operation and maintenance service affects the operation efficiency of enterprises.To realize the intelligent service desk based on unattended,a semantic analysis method of IT operation and maintenance service requirements was proposed,including two tasks of intention recognition and named entity recognition.Based on the Multi-BERT-BiLSTM-CRF(MBBC)benchmark model,the part-of-speech and entity dictionary features were integrated into the coding layer with prior knowledge and external resources to enhance the learning of lexical information and domain knowledge.In addition,the parameter sharing mode of MBBC model was improved,and Enhanced MBBC(EMBBC)model was proposed to enhance the information sharing capability between two tasks.The computational experiments on an enterprise IT operation and maintenance worksheet data set showed that compared with MBBC model,the recognition performance of the two tasks could be further improved by combining the features of speech and entity dictionary and adopting EMBBC model.

关 键 词：IT运维服务 意图识别 命名实体识别 BERT模型 多任务学习 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]