多模型融合的客服工单文本分类方法的研究与实现

doi:10.11959/j.issn.1000-0801.2021236

电信科学 ›› 2021, Vol. 37 ›› Issue (11): 86-96.doi: 10.11959/j.issn.1000-0801.2021236

多模型融合的客服工单文本分类方法的研究与实现

张亮¹, 代晓菊¹, 郑荣¹, 贺同泽²

¹ 上海理想信息产业（集团）有限公司，上海 201315
² 北京邮电大学，北京 100876

修回日期:2021-10-14 出版日期:2021-11-20 发布日期:2021-11-01
作者简介:张亮（1991− ），男，上海理想信息产业（集团）有限公司软件产品研发工程师，主要研究方向为人工智能技术、自然语言处理、大数据挖掘与分析
代晓菊（1990− ），女，上海理想信息产业（集团）有限公司软件产品研发工程师，主要研究方向为人工智能技术、自然语言处理、大数据挖掘与分析
郑荣（1981− ），男，上海理想信息产业（集团）有限公司软件产品研发高级工程师，主要研究方向为人工智能技术、自然语言处理、大数据挖掘与分析
贺同泽（1996− ），男，北京邮电大学硕士生，主要研究方向为推荐系统、自然语言处理、大数据挖掘与分析
基金资助:
上海互联网大数据工程技术研究中心资助项目(15DZ2250700)

Research and implementation of text classification method for customer service orders based on multi-model fusion

Liang ZHANG¹, Xiaoju DAI¹, Rong ZHENG¹, Tongze He²

¹ Shanghai Ideal Information Industry (Group) Co., Ltd., Shanghai 201315, China
² Beijing University of Posts and Telecommunications, Beijing 100876, China

Revised:2021-10-14 Online:2021-11-20 Published:2021-11-01
Supported by:
Shanghai Internet Big Data Engineering Technology Research Center(15DZ2250700)

摘要/Abstract

摘要：

电信呼叫中心客服在人工进行工单分类时存在归档耗时长、效率低、准确率难以保障的问题，但此场景下类别数量多，且类别间具有层级关联，导致传统文本分类方法准确率较低。针对此问题，提出了一种基于多模型融合的文本分类方法，根据不同层级的数据特点使用不同模型进行分类，考虑了类别的层级关联以提升准确率，并验证了此方法的有效性，可以优化客服生产系统运营流程，加快现场人工客服响应能效，提升客服热线整体运营效率，实现人工智能注智生产。

关键词: 文本分类, 客服工单, 多模型融合, 运营效率

Abstract:

Due to the large amount of order categories and their hierarchical associations, traditional manual order classification method of customer service in telecom call center has the problems of long archiving time, low efficiency and unsustainable accuracy.To solve this problem, a novel text classification algorithm based on multi-model fusion was proposed, which intelligently classify orders with multiple models based on data characteristics and their hierarchical associations, the effectiveness of this method was verified.The current manual operation process was optimized and operation efficiency was enhanced, which support the intelligent transformation and upgradation of existing customer service system.

Key words: text classification, customer service order, multi-model fusion, operational efficiency

中图分类号:

TP391.1

张亮, 代晓菊, 郑荣, 贺同泽. 多模型融合的客服工单文本分类方法的研究与实现[J]. 电信科学, 2021, 37(11): 86-96.

Liang ZHANG, Xiaoju DAI, Rong ZHENG, Tongze He. Research and implementation of text classification method for customer service orders based on multi-model fusion[J]. Telecommunications Science, 2021, 37(11): 86-96.

图/表 11

图1

图2

图3

表1

图4

图5

图6

图7

表2

图8

图9

参考文献 9

[1]	KIM Y , . Convolutional neural networks for sentence classification[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP). USA:Association for Computational Linguistics, 2014.
[2]	MA M B , HUANG L , ZHOU B W ,et al. Dependency-based convolutional neural networks for sentence embedding[C]// Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2:Short Papers). USA:Association for Computational Linguistics, 2015.
[3]	Devlin J , Chang M W , Lee K ,et al. BERT:pre-training of deep bidirectional transformers for language understanding[EB]. 2018.
[4]	ZHAO Z W , WU Y Z . Attention-based convolutional neural networks for sentence classification[C]// Interspeech 2016.[S.l:s.n.], 2016.
[5]	SHIMURA K , LI J Y , FUKUMOTO F . HFT-CNN:learning hierarchical category structure for multi-label short text categorization[C]// Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. USA:Association for Computational Linguistics, 2018.
[6]	CHEN T Q , GUESTRIN C . XGBoost:a scalable tree boosting system[C]// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM Press, 2016: 785-794.
[7]	ZHANG Z Y , HAN X , LIU Z Y ,et al. ERNIE:enhanced language representation with informative entities[C]// Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. USA:Association for Computational Linguistics, 2019.
[8]	KALCHBRENNER N , GREFENSTETTE E , BLUNSOM P . A convolutional neural network for modelling sentences[C]// Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1:Long Papers). USA:Association for Computational Linguistics, 2014.
[9]	YIN W P , SCHüTZE H ,, . Multichannel variable-size convolution for sentence classification[C]// Proceedings of the Nineteenth Conference on Computational Natural Language Learning. USA:Association for Computational Linguistics, 2015.

层级	类别数	平均每个类别数据量	TextCNN准确率
第一级	9	8 942	0.9
第二级	68	1 183	0.819
第三级	256	314	0.674
第四级	654	123	0.601

模型	预处理	第一级	第二级	第三级	第四级	最终
TextCNN	分词+词向量	—	—	—	—	0.601
分级TextCNN	分词+词向量	0.901	0.910	0.902	0.841	0.622
BERT	词向量	—	—	—	—	0.619
分级BERT	词向量	0.893	0.903	0.884	0.907	0.647
XGBoost	特征提取	—	—	—	—	0.582
分级XGBoost	特征提取	0.867	0.880	0.912	0.916	0.634
HFT-CNN	分词+词向量	0.898	0.908	0.886	0.894	0.646
MMF	分词+词向量+特征提取	0.901	0.910	0.912	0.916	0.685

多模型融合的客服工单文本分类方法的研究与实现

Research and implementation of text classification method for customer service orders based on multi-model fusion

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 9

相关文章 6

Metrics

推荐阅读 0

[1]	宋勇, 严志伟, 秦玉坤, 赵东明, 叶晓舟, 柴园园, 欧阳晔. 基于矩阵分解和注意力多任务学习的客服投诉工单分类[J]. 电信科学, 2022, 38(2): 103-110.
[2]	赵进, 杨小军. 基于GRW和FastText模型的电信用户投诉文本分类应用[J]. 电信科学, 2021, 37(6): 125-131.
[3]	汪少敏,杨迪,任华. 基于深度学习的文本分类系统关键技术研究与模型验证[J]. 电信科学, 2018, 34(12): 117-124.
[4]	寇晓淮,程华. 基于主题模型的垃圾邮件过滤系统的设计与实现[J]. 电信科学, 2017, 33(11): 73-82.
[5]	陈康,付华峥,陈翀,张宇艺,韦誉. 基于DPI的用户兴趣实时分类[J]. 电信科学, 2016, 32(12): 109-115.
[6]	边凌燕,贺仁龙,姚晓辉. 基于DPI数据挖掘实现URL分类挂载的相关技术研究[J]. 电信科学, 2013, 29(11): 6-11.