Cross-domain Text Sentiment Classification Based on Auxiliary Classification Networks

doi:10.16182/j.issn1004731x.joss.21-1283

Abstract

Abstract:

To align exactly the texts with same sentiment polarities of source and target domains, and to enlarge the feature difference of different sentiment texts as much as possible, a domain adaptation model with weighted adversarial networks is proposed. A new structured classification network consisting of a main classification network and an auxiliary classification network is proposed, in which the main classification network is used to perform supervised learning on the labeled texts of the source domain, and the auxiliary classification network is used to improve the distinguishability of the text features. A calculation method of multiple adversarial network weights is proposed to realize the exact alignment of same class samples of different domains. Experimental results show that, for Amazon dataset, the average recognition accuracy for the texts of target domains can reach 84.22%, which is 2.07% higher than the compared models. The optimized feature extractor and the feature classifier can be applied to the source and target domains simultaneously on the proposed model, and can provide reliable data for the simulation and modeling of text analysis in different fields.

Key words: text sentiment classification, domain adaptation, adversarial network, auxiliary classification network

CLC Number:

TP391.4

Na Ma, Tingxin Wen, Xu Jia, Xiaohui Li. Cross-domain Text Sentiment Classification Based on Auxiliary Classification Networks[J]. Journal of System Simulation, 2023, 35(4): 721-733.

Figures/Tables 14

Fig. 1

Fig. 2

Fig. 3

Table 1

Table 2

Recognition results under different parameters

$λ d$	$λ c$
$λ d$	0.1	0.5	1.0	5.0	10.0
0.1	0.774	0.795	0.821	0.814	0.769
0.5	0.814	0.826	0.832	0.792	0.788
1.0	0.768	0.796	0.812	0.785	0.772
5.0	0.744	0.785	0.776	0.765	0.748
10.0	0.759	0.779	0.751	0.771	0.749

Table 2

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Table 3

Comparison of recognition results of target domains when using different algorithms

源域→目标域	SVM	AuxNN	DANN	PBLM	DAS	ACAN	本文
平均值	73.66	79.85	80.29	80.40	81.96	82.15	84.22
$D → B$	75.20	80.80	81.70	82.50	82.05	82.35	82.41
$E → B$	68.85	78.00	78.55	71.40	80.00	79.75	82.57
$K → B$	70.00	77.85	79.25	74.20	80.05	80.80	82.54
$B → D$	77.15	81.75	82.30	84.20	82.75	83.45	85.36
$E → D$	69.50	80.65	79.70	75.00	80.15	81.75	82.82
$K → D$	71.40	78.90	80.45	79.80	81.40	82.10	83.32
$B → E$	72.15	76.40	77.60	77.60	81.15	81.30	81.26
$D → E$	71.65	77.55	79.70	79.60	81.55	82.80	84.88
$K → E$	79.75	84.05	86.65	87.10	85.80	86.60	89.13
$B → K$	73.50	78.10	76.10	82.50	82.25	83.05	84.72
$D → K$	72.00	80.05	77.35	83.20	81.50	78.60	82.89
$E → K$	82.80	84.15	83.95	87.80	84.85	83.35	88.76

Table 3

Fig. 10

Table 4

User-item interest analysis based on review text

物品	$u 1$	$u 2$
手机 $i 1$ 评论文本	打游戏不卡顿，画面流畅	手机续航能力强
书籍 $i 2$ 评论文本	章节安排不是很合理	存在错别字，图印刷不清晰
台式机 $i 3$ 评论文本	处理器与显卡性能好	系统运行非常稳定，从不死机
游戏笔记本 $i 4$ 兴趣度	高	？
商务笔记本 $i 5$ 兴趣度	低	？

Table 4

References 31

1	赵传君, 王素格, 李德玉. 跨领域文本情感分类研究进展[J]. 软件学报, 2020, 31(6): 1723-1726.
	Zhao Chuanjun, Wang Suge, Li Deyu. Research Progress on Cross-Domain Text Sentiment Classification[J]. Journal of Software, 2020, 31(6): 1723-1746.
2	Robert R. Domain Adaptation Using Domain Similarity and Domain Complexity-based Instance Selection for Cross-Domain Sentiment Analysis[C]//IEEE 12th International Conference on Data Mining Workshops. Piscataway: IEEE, 2012: 717-723.
3	Xia R, Zong C Q, Hu X L, et al. Feature Ensemble Plus Sample Selection: Domain Adaptation for Sentiment Classification[J]. IEEE Intelligent Systems(S1541-1672), 2013, 28(3): 10-18.
4	Giacomo D, Gianluca M, Andrea P. Cross-Domain Sentiment Classification via Polarity-Driven State Transitions in a Markov Model[C]//International Joint Conference on Knowledge Discovery, Knowledge Engineering, and Knowledge Management. Berlin: Springer, 2015: 118-138.
5	John B, Ryan T M, Fernando P. Domain Adaptation with Structural Correspondence Learning[C]//Conference on Empirical Methods in Natural Language Processing. New York: ACM, 2006: 120-128.
6	Yftah Z, Roi R. Neural Structural Correspondence Learning for Domain Adaptation[C]//21st Conference on Computational Natural Language Learning. Stroudsburg: ACL, 2017: 400-410.
7	Wang L, Niu J W, Song H B, et al. SentiRelated: A Cross-Domain Sentiment Classification Algorithm for Short Texts through Sentiment Related Index[J]. Journal of Network and Computer Applications(S1084-8045), 2018, 101: 111-119.
8	Tareq A M, Nazlia O, Mohammed A, et al. Feature Transfer through New Statistical Association Measure for Cross-Domain Sentiment Analysis[J]. Journal of Engineering and Applied Sciences(S1816-949X), 2017, 12(1): 164-170.
9	Zhu E D, Huang G Y, Mo B Y, et al. Features Extraction Based on Neural Network for Cross-Domain Sentiment Classification[C]//International Conference on Database Systems for Advanced Applications. Berlin: Springer, 2016: 81-88.
10	Ganin Y, Lempitsky V. Unsupervised Domain Adaptation by Backpropagation[C]//32nd International Conference on Machine Learning. Cambridge: MIT Press, 2015: 1180-1189.
11	Li Z, Zhang Y, Wu Y X, et al. End-to-End Adversarial Memory Network for Cross-Domain Sentiment Classification[C]//International Joint Conference on Artificial Intelligence. New York: ACM, 2017: 2237-2243.
12	Qu X Y, Zou Z K, Cheng Y, et al. Adversarial Category Alignment Network for Cross-Domain Sentiment Classification[C]//2019 Conference of the North American Chapter of the Association for Computational Linguistics. Stroudsburg: ACL, 2019: 2496-2508.
13	He R D, Wee S L, Hwee T N, et al. Adaptive Semi-Supervised Learning for Cross-Domain Sentiment Classification[C]//2018 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2018: 3467-3476.
14	Zhao C J, Wang S G, Li D Y. Multi-Source Domain Adaptation with Joint Learning for Cross-Domain Sentiment Classification[J]. Knowledge-Based Systems(S0950-7051), 2020, 191: 105254.
15	Li L, Ye W R, Long M S, et al. Simultaneous Learning of Pivots and Representations for Cross-Domain Sentiment Classification[C]//AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2020: 8220-8227.
16	Mao K L, Niu J W, Wang X J, et al. Cross-Domain Sentiment Analysis of Product Reviews by Combining Lexicon-Based and Learn-Based Techniques[C]//IEEE 17th International Conference on High Performance Computing and Communications. Piscataway: IEEE, 2015: 351-356.
17	Dolores M G, Eugenio M C, Teresa M V, et al. Cross-Domain Sentiment Analysis Using Spanish Opinionated Words[C]//International Conference on Applications of Natural Language to Data Bases/Information Systems. Berlin: Springer, 2014: 214-219.
18	Huang X C, Rao Y H, Xie H R, et al. Cross-Domain Sentiment Classification via Topic-Related TrAdaBoost[C]//Thirty-First AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2017: 4939-4940.
19	Zhou G Y, Zhou Y, Guo X Y, et al. Cross-Domain Sentiment Classification via Topical Correspondence Transfer[J]. Neurocomputing(S0925-2312), 2015, 159: 298-305.
20	Giacomo D, Gianluca M, Andrea P, et al. Markov Chain Based Method for In-Domain and Cross-Domain Sentiment Classification[C]//7th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management. Piscataway: IEEE, 2015: 127-137.
21	Nelakurthi A R, Tong H h, Maciejewski R, et al. User-Guided Cross-Domain Sentiment Classification[C]//2017 SIAM International Conference on Data Mining. Bethesda: SIAM, 2017: 471-479.
22	Pei Z Y, Cao Z J, Long M S, et al. Multi-Adversarial Domain Adaptation[C]//Thirty-second AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2018: 3934-3941.
23	John B, Mark D, Fernando P. Biographies, Bollywood, Boom-Boxes and Blenders: Domain Adaptation for Sentiment Classification[C]//The 45th Annual Meeting of the Association of Computational Linguistics. Stroudsburg: ACL, 2007: 440-447.
24	Chen M M, Xu Z X, Kilian Q W, et al. Marginalized Denoising Autoencoders for Domain Adaptation[C]//29th International Coference on International Conference on Machine Learning. New York: ACM, 2012: 1627-1634.
25	Ganin Y, Ustinova E, Lempitsky V, et al. Domain-Adversarial Training of Neural Networks[J]. Journal of Machine Learning Research(S1532-4435), 2016, 17: 1-35.
26	Yu J F, Jiang J. Learning Sentence Embeddings with Auxiliary Tasks for Cross-Domain Sentiment Classification[C]//2016 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2016: 236-246.
27	Yftah Z, Roi R. Pivot Based Language Modeling for Improved Neural Domain Adaptation[C]//2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2018: 1241-1251.
28	He R D, Lee W S, Hwee T N, et al. Adaptive Semi-Supervised Learning for Cross-Domain Sentiment Classification[C]//2018 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2018: 3467-3476.
29	刘群, 孙红涛, 纪良浩. 一种融合文本语义和情感分析的好友推荐方法[J]. 系统仿真学报, 2016, 28(11): 2852-2859.
	Liu Qun, Sun Hongtao, Ji Lianghao. Friend Recommendation Based on Analysis of Users' Emotions and Text Semantics[J]. Journal of System Simulation, 2016, 28(11): 2852-2859.
30	庄穆妮, 李勇, 谭旭, 等. 基于BERT-LDA模型的新冠肺炎疫情网络舆情演化仿真[J]. 系统仿真学报, 2021, 33(1): 24-36.
	Zhuang Muni, Li Yong, Tan Xu, et al. Evolutionary Simulation of Online Public Opinion Based on the BERT-LDA Model under COVID-19[J]. Journal of System Simulation, 2021, 33(1): 24-36.
31	Gao K, Anandhan P, Kumar R. Analysis and Evaluation of the Regional Air Quality Index Forecasting Based on Web-Text Sentiment Analysis Method[J]. Environmental Impact Assessment Review(S0915-9255), 2021, 87: 106514.

域	样本数目	正样本示例	负样本示例
B	4 465	这是一本非常棒的书，老少皆宜，它以一种有趣的方式展示了为什么我们需要一种通用的测量形式	这本书传达了一个很好的信息，可能会对我产生持久的影响，但情节非常乏味，书的中间部分基因和鳍鱼的关系似乎与书的其余部分不相适应，不推荐
D	3 586	如果你没有这张DVD，你需要把它添加到你的收藏中。在我看来，这是有史以来最好的美国动画电影	对于那些对这部电影赞不绝口的人来说，你一定是在另一个维度或别的什么地方
E	5 681	我非常喜欢它，我把它用在野外狩猎的户外小径相机里，它能拍几张照片，在各种天气的相机里，我还会再买金斯顿	我最近刚买了这个，当我安装它时，诺顿在安装光盘上捕获了2个病毒
K	5 945	我很喜欢用这个铸造锅，它煮得很好吃，容量很大	绝对的劣质产品，不要浪费你的钱，当你磨任何东西的时候，它都不会固定住，这个产品肯定会让你发脾气

[1]	Xingquan Cai, Zhijun Li, Mengyao Xi, Haiyan Sun. Costume Pattern Sketch Colorization and Style Transfer Based on Neural Network [J]. Journal of System Simulation, 2023, 35(3): 604-615.
[2]	Jianfang Cao, Yiming Jia, Minmin Yan, Xiaodong Tian. Murals Super-resolution Reconstruction with the Stable Enhanced Generative Adversarial Network [J]. Journal of System Simulation, 2022, 34(5): 1076-1089.
[3]	Xu Kang, Xiaofeng Zhang. Radar Remote Sensing Data Augmentation Method Based on Generative Adversarial Network [J]. Journal of System Simulation, 2022, 34(4): 920-927.
[4]	Hu Lei, Wang Zugen, Chen Tian, Zhang Yongmei. An Improved SRGAN Infrared Image Super-Resolution Reconstruction Algorithm [J]. Journal of System Simulation, 2021, 33(9): 2109-2118.
[5]	Cheng Wencong, Shi Xiaokang, Wang Zhigang. Creating Synthetic Satellite Cloud Data Based on GAN Method [J]. Journal of System Simulation, 2021, 33(6): 1297-1306.
[6]	Li Xinli, Zou Changming, Yang Guotian, Liu He. Research of Super-resolution Processing of Invoice Image Based on Generative Adversarial Network [J]. Journal of System Simulation, 2021, 33(6): 1307-1314.