[Objective] This paper analyzes online comments by professional critics and average audience, aiming to improve the sentiment classification of reviews. [Methods] First, we introduced the professional backgrounds of contributors to examine the emotional polarity of reviews. Then, we used the generative adversarial network to decide whether the contributor was a professional critic or an average browser. Finally, we identified their differences to further improve the accuracy of emotion classification. [Results] The accuracy rate of the proposed model reached 83.6%, which was 5.6% higher than the benchmark model LSTM and 4.4% higher than BiLSTM. [Limitations] We only studied movie reviews, and more research is needed to evaluate our model with data sets from other fields. [Conclusions] The proposed GJOINT model can effectively improve the results of sentiment classification of online reviews.
Funny and charming, it’s a very good movie. (既有趣又迷人,是一部很好的电影。)
[E2]
This version of Black Panther weaves a seemingly disparate set of strands together rather seamlessly, presenting more than enough comic book acrobatics and mayhem to satisfy the core fan base while also speaking a necessary degree of truth to power. (这个版本的黑豹把一组看似不同的线编织在一起,展现了足够多的漫画书,杂技和混乱,以满足核心粉丝群,同时也展示了必要的真实性。)
Table 1 评论实例
Fig.1 情感分类和评论专业类型分类联合模型
Fig.2 基于生成式对抗网络和评论专业类型的情感分类模型
实验模型
实验一准确率
实验二准确率
A
C
A
C
SingleCategory
0.678
0.734
0.768
0.802
CrossCategory
0.648
0.623
0.746
0.706
MultiCategory
0.712
0.685
0.782
0.816
Table 2 专业类型不同的评论情感分类结果
模型名称
实验一准确率
实验二准确率
LSTM
0.687
0.780
BiLSTM
0.672
0.792
JOINT
0.703
0.806
GJOINT
0.716
0.836
Table 3 情感分类性能对比
评论内容
LSTM
GJOINT
[E3] The most courageous thing about it, from today’s standards, is that it closes without an obligatory happy ending, and an audience that has watched for 187 minutes doesn’t get a tidy, mindless conclusion. (从今天的标准来看,它最有勇气的一点是结束时没有一个必然的快乐结局,一个观看了187分钟的观众不会得到一个整洁、愚蠢的结论。)
Negative
Positive
[E4] boring old hollywood style epic with that canned elevator music. still it was the best of its kind. (无聊的老好莱坞风格史诗与罐装电梯音乐。尽管如此,它还是同类型中最好的。)
Negative
Positive
Table 4 GJOINT模型有效性评估实例
[1]
Wen X Z, Shao L, Xue Y, et al. A Rapid Learning Algorithm for Vehicle Classification[J]. Information Sciences, 2015,295(C):395-406.
doi: 10.1016/j.ins.2014.10.040
[2]
Chen B J, Shu H Z, Coatrieux G, et al. Color Image Analysis by Quaternion-Type Moments[J]. Journal of Mathematical Imaging and Vision, 2015,51(1):124-144.
doi: 10.1007/s10851-014-0511-6
[3]
Gu B, Sheng V S, Wang Z J, et al. Incremental Learning for ν-Support Vector Regression[J]. Neural Networks, 2015,67:140-150.
doi: 10.1016/j.neunet.2015.03.013
( Li Hui, Chai Yaqing. Fine-Grained Sentiment Analysis Based on Convolutional Neural Network[J]. Data Analysis and Knowledge Discovery, 2019,3(1):95-103.)
[5]
Tang D Y, Qin B, Liu T. Document Modeling with Gated Recurrent Neural Network for Sentiment Classification[C]// Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing. 2015: 1422-1432.
[6]
Wang S D, Manning C D. Baselines and Bigrams: Simple, Good Sentiment and Topic Classification[C]// Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics. 2012: 90-94.
[7]
Song G, Ye Y M, Du X L, et al. Short Text Classification: A Survey[J]. Journal of Multimedia, 2014,9(5):635-643.
( Zeng Ziming, Yang Qianwen. Sentiment Analysis for Micro-blogs with LDA and AdaBoost[J]. Data Analysis and Knowledge Discovery, 2018,2(8):51-59.)
[9]
Hu F, Li L, Zhang Z L, et al. Emphasizing Essential Words for Sentiment Classification Based on Recurrent Neural Networks[J]. Journal of Computer Science and Technology, 2017,32(4):785-795.
doi: 10.1007/s11390-017-1759-2
( Zhao Yanyan, Qin Bing, Liu Ting. Sentiment Analysis[J]. Journal of Software, 2010,21(8):1834-1848.)
[11]
Tai K S, Socher R, Manning C D. Improved Semantic Representations from Tree-Structured Long Short-Term Memory Networks[OL]. arXiv Preprint, arXiv:1503.00075v2.
( Wu Peng, Ying Yang, Shen Si. Negative Emotions of Online Users’ Analysis Based on Bidirectional Long Short-Term Memory[J]. Journal of the China Society for Scientific and Technical Information, 2018,37(8):845-853.)
[13]
Li X, Bing L D, Lam W, et al. Transformation Networks for Target-Oriented Sentiment Classification[C]// Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics. 2018: 946-956.
( Lu Qiang, Zhu Zhenfang, Xu Fuyong, et al. Chinese Sentiment Classification Method with Bi-LSTM and Grammar Rules[J]. Data Analysis and Knowledge Discovery, 2019,3(11):99-107.)
[15]
Hovy D. Demographic Factors Improve Classification Performance[C]// Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference of the Asian Federation of Natural Language Processing. 2015: 752-762.
[16]
Vosoughi S, Zhou H, Roy D, et al. Enhanced Twitter Sentiment Classification Using Contextual Information[C]// Proceedings of the 6th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis. 2015: 16-24.
( Zhao Dongmei, Li Ya, Tao Jianhua, et al. Sentiment Analysis Based on Collaborative Filter Attention Mechanism[J]. Journal of Chinese Information Processing, 2018,32(8):128-134.)
[18]
Hochreiter S, Schmidhuber J. Long Short-Term Memory[J]. Neural Computation, 1997,9(8):1735-1780.
pmid: 9377276
[19]
Wang K, Wan X J. SentiGAN: Generating Sentimental Texts via Mixture Adversarial Networks[C]// Proceedings of the 27th International Joint Conference on Artificial Intelligence. 2018: 4446-4452.
[20]
William F, Ian G, Andrew M D, et al. MaskGAN: Better Text Generation via Filling in the[C]// Proceedings of the 6th International Conference on Learning Representations. 2018: 1-17.