基于特征融合和集成学习的建议语句分类模型

doi:10.6040/j.issn.1672-3961.0.2018.207

摘要/Abstract

摘要：

建议挖掘作为一项新兴研究任务近年来逐渐受到了研究者的关注。与英文相比,中文的建议表达形式更为丰富,呈现出许多不同特点,因此有必要在中文环境下开展建议挖掘研究。针对建议挖掘中的建议语句检测这一核心任务,提出一种综合应用Stacking和Bagging方法的集成学习模型来进行建议语句分类。使用Stacking组合分类器来构建概率特征空间,分别使用卷积神经网络(convolutional neural network, CNN)和段落向量模型(paragraph vector, PV)构建评论文本的CNN特征空间和段落向量特征空间,对上述特征进行融合,并训练Bagging分类器来对建议语句分类。在中文数据集上的试验结果验证了本研究模型的有效性。

关键词: 建议挖掘, 建议语句分类, 卷积神经网络, 集成学习, 特征融合

Abstract:

As an emerging research task, suggestion mining has gradually attracted attention of researchers in recent years. Compared with English language suggestion expression forms, those of Chinese were more abundant, and many different characteristics were present. It was necessary to carry out the research on suggestion mining in the Chinese environment. As suggestion sentence detection was the core task of suggestion mining, this research proposed an ensemble learning model that integrated the Stacking and Bagging methods to classify the reviews for the detection of suggestion sentence. The model firstly used Stacking to combine classifiers and constructed probabilistic feature space. Then, the convolution neural network (CNN) and paragraph vector (PV) model were used to construct the CNN feature space and paragraph vector feature space of the reviews respectively. Finally, the above features were fused and the Bagging classifier was trained to classify suggestion sentences. Experimental results on Chinese dataset verified the effectiveness of the model.

Key words: suggestion mining, suggestion sentence classification, convolutional neural network, ensemble learning, feature fusion

中图分类号:

TP391.1

张璞,刘畅,王永. 基于特征融合和集成学习的建议语句分类模型[J]. 山东大学学报 (工学版), 2018, 48(5): 47-54.

Pu ZHANG,Chang LIU,Yong WANG. Suggestion sentence classification model based on feature fusion and ensemble learning[J]. Journal of Shandong University(Engineering Science), 2018, 48(5): 47-54.

图/表 6

图1

图2

图3

表1

表2

表3

参考文献 20

1	赵妍妍, 秦兵, 刘挺. 文本情感分析[J]. 软件学报, 2010, 21 (8): 1834- 1848.
	ZHAO Yanyan , QIN Bing , LIU Ting . Sentiment analysis[J]. Journal of Software, 2010, 21 (8): 1834- 1848.
2	李然, 林政, 林海伦, 等. 文本情绪分析综述[J]. Journal of Computer Research and Development, 2018, (55): 30- 52.
	LI Ran , LIN Zheng , LIN Hailun , et al. Text emotion analysis: a survey[J]. Journal of Computer Research and Development, 2018, (55): 30- 52.
3	刘兵.情感分析:挖掘观点、情感和情绪[M].北京:机械工业出版社, 2017.07.
4	NEGI S. Suggestion mining from opinionated text[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics—Student Research Workshop. Association for Computational Linguistics. Stroudsburg, USA: ACL, 2016: 7-12.
5	RAMANAND J, BHAVSAR K, PEDANEKAR N. Wishful thinking: finding suggestions and 'buy' wishes from product reviews[C]//Proceedings of the NAACL HLT 2010 Workshop on Computational Approaches to Analysis and Generation of Emotion in Text. Stroudsburg, USA: ACL, 2010: 54-61.
6	BRUN C , HAGEGE C . Suggestion mining: detecting suggestions for improvement in users' comments[J]. Research in Computing Science, 2013, (70): 199- 209.
7	NEGI S, BUITELAAR P. Curse or boon? presence of subjunctive mood in opinionated text[C]//Proceedings of the 11th International Conference on Computational Semantics. Stroudsburg, USA: ACL, 2015: 101-106.
8	WICAKSONO A F, MYAENG S H. Automatic extraction of advice-revealing sentences for advice mining from online forums[C]//International Conference on Knowledge Capture. New York, USA: ACM, 2013: 97-104.
9	DONG Li, WEI Furu, DUAN Yajuan, et al. The automated acquisition of suggestions from tweets[C]//Proceedings of the Twenty-Seventh American Association for Artificial Intelligence. Menlo Park, Canada: AAAI, 2013: 239-245.
10	LAI Siwei, XU Liheng, LIU Kang, et al. Recurrent convolutional neural networks for text classification[C]//Twenty-Ninth AAAI Conference on Artificial Intelligence. Texas Austin, USA: AAAI, 2015: 2267-2273.
11	YANG Zichao, YANG Diyi, DYER C, et al. Hierarchical attention networks for document classification[C]//Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. SAN Diego, USA: NAACL, 2016: 1480-1489.
12	NEGI S, ASOOJA K, MEHROTRA S, et al. A study of suggestions in opinionated texts and their automatic detection[C]//Proceedings of the Fifth Joint Conference on Lexical and Computational Semantics. Stroudsburg, USA: ACL, 2016: 170-178.
13	孙松涛, 何炎祥. 基于CNN特征空间的微博多标签情感分类[J]. 工程科学与技术, 2017, 49 (3): 162- 169.
	SUN Songtao , HE Yanxiang . Multi-label emotion classification for microblog based on CNN feature space[J]. Advanced Engineering Sciences, 2017, 49 (3): 162- 169.
14	LE Q, MKOLOV T. Distributed representations of sentences and documents[C]//Proceedings of the 31st International Conference on Machine Learning. Beijing, China: ICML, 2014: 1188-1196.
15	李寿山, 黄居仁. 基于Stacking组合分类方法的中文情感分类研究[J]. 中文信息学报, 2010, 24 (5): 56- 61. doi: 10.3969/j.issn.1003-0077.2010.05.010
	LI Shoushan , HUANG Juren . Chinese sentiment classification based on stacking combination method[J]. Journal of Chinese Information Processing, 2010, 24 (5): 56- 61. doi: 10.3969/j.issn.1003-0077.2010.05.010
16	李恒超, 林鸿飞, 杨亮, 等. 一种用于构建用户画像的二级融合算法框架[J]. 计算机科学, 2018, 45 (1): 157- 161.
	LI Hengchao , LIN Hongfei , YANG Liang , et al. Two-level stacking algorithm framework for building user portrait[J]. Computer Science, 2018, 45 (1): 157- 161.
17	KIM Y. Convolutional neural networks for sentence classification[C]//Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, USA: ACL, 2014: 1746-1751.
18	NGUYEN T H, GRISHMAN R. Relation extraction: perspective from convolutional neural networks[C]//Proceedings of the NAACL Workshop on Vector Space Modeling for NLP. Denver Colorado, Canada: NAACL, 2015: 39-48.
19	CHENG T, GUESTRIN C. Xgboost: a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD Inernational Conference on Knowledge Discovery and Data Mining. New York, USA: ACM, 2016: 785-794.
20	三星电子.三星盖乐世社区产品建议论坛[EB/OL]. [2018-07-23]. http://www.galaxyclub.cn/bbs/productadvicearea/productadvice.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed

%
模型	精确率	召回率	F值	准确率
NB	83.42	81.11	82.25	82.17
FM	87.03	81.17	84.00	84.23
LR	88.58	82.05	85.19	85.20
RF	87.36	82.55	84.89	85.01
ET	86.71	83.02	84.82	84.87
TextCNN	88.14	72.10	79.32	79.70
Pos-TextCNN	84.37	79.62	81.93	81.49
Stacking+Bagging	87.48	85.20	86.32	86.25
Stacking+Bagging+CNN	87.92	84.90	86.38	86.47
Stacking+Bagging+PV	88.27	85.14	86.68	86.66
CNN+PV+Bagging	86.86	82.97	84.87	84.87
Stacking+Bagging+CNN+PV	88.63	86.06	87.33	87.28

序号	评论文本	NB	FM	LR	RF	ET	S-B	P	T
1	我的9350手机升级后2天,屏幕右有条红线怎么处理,希望大神回复。	0	0	0	1	1	0	0	0
2	小小国家sx公司太欺负中国人了,我以后永远都不买他们任何一件产品,我建议全中国人都别再买,爱我中华,支持国产	0	0	1	1	1	0	0	0
3	三星c7系统更新后变成砖头了,一个月前买的三星c7当时续航能力还不错,最近系统更新后让他名不副实,原来充电一小时能充满现在充两小时都充不满而且耗电非常快,问客服说建议恢复出厂设置或者关机充电。	0	0	1	1	1	0	0	0
4	三星都有那几款,准备花4 000~5 500之间买个三星,求建议!	1	1	1	0	0	1	0	0

序号	评论文本	NB	FM	LR	RF	ET	S-B	P	T
1	关于指纹解锁不灵敏问题,指纹解锁实在太不灵敏了,平常都不敢开启双击启动相机。和ip6s差几个档次,和n5机皇定位严重不符。	1	1	1	0	0	1	0	0
2	三星下一代手机应该具有的特性(一个都不要少), 1.手写笔, 2.红外线遥控, 3.防水(可以游泳级别), 4.高清屏(4K,不玩VR没感受,玩了就知道了)	1	1	0	0	0	0	1	1
3	给C7更新Grace UX系统吧,很喜欢这个新的定制系统,简洁易用, C7出的时间也不长,不能不管C7。	0	0	0	0	0	0	1	1

[1]	沈冬冬,周风余,栗梦媛,王淑倩,郭仁和. 基于集成深度神经网络的室内无线定位[J]. 山东大学学报 (工学版), 2018, 48(5): 95-102.
[2]	梁蒙蒙,周涛,夏勇,张飞飞,杨健. 基于PSO-ConvK卷积神经网络的肺部肿瘤图像识别[J]. 山东大学学报 (工学版), 2018, 48(5): 77-84.
[3]	何正义,曾宪华,郭姜. 一种集成卷积神经网络和深信网的步态识别与模拟方法[J]. 山东大学学报(工学版), 2018, 48(3): 88-95.
[4]	谢志峰,吴佳萍,马利庄. 基于卷积神经网络的中文财经新闻分类方法[J]. 山东大学学报(工学版), 2018, 48(3): 34-39.
[5]	赵彦霞, 王熙照. 基于SVD和DCNN的彩色图像多功能零水印算法[J]. 山东大学学报(工学版), 2018, 48(3): 25-33.
[6]	牟春倩,唐雁. 融合整体和局部信息的三维模型检索方法[J]. 山东大学学报(工学版), 2016, 46(6): 48-53.
[7]	王斌,常发亮,刘春生. 基于多特征融合的交通标志分类[J]. 山东大学学报(工学版), 2016, 46(4): 34-40.
[8]	王立宏,李强. 旅行商问题的一种选择性集成求解方法[J]. 山东大学学报(工学版), 2016, 46(1): 42-48.
[9]	陈大伟,闫昭*,刘昊岩. SVD系列算法在评分预测中的过拟合现象[J]. 山东大学学报(工学版), 2014, 44(3): 15-21.
[10]	孔超1,2,张化祥1,2*,刘丽1,2. 基于兴趣区域特征融合的半监督图像检索算法[J]. 山东大学学报(工学版), 2014, 44(3): 22-28.
[11]	徐姗姗,刘应安*,徐昇. 基于卷积神经网络的木材缺陷识别[J]. 山东大学学报(工学版), 2013, 43(2): 23-28.
[12]	房晓南1,2,张化祥1,2*,高爽1,2. 基于SMOTE和随机森林的Web spam检测[J]. 山东大学学报(工学版), 2013, 43(1): 22-27.
[13]	张伶卫,万文强. 基于云计算平台的代价敏感集成学习算法研究[J]. 山东大学学报(工学版), 2012, 42(4): 19-23.
[14]	谢伙生,刘敏. 一种基于主动学习的集成协同训练算法[J]. 山东大学学报(工学版), 2012, 42(3): 1-5.
[15]	李小斌1, 李世银2. 时间序列早期分类的多分类器集成方法[J]. 山东大学学报(工学版), 2011, 41(4): 73-78.