结合语言知识和深度学习的中文文本情感分析方法

doi:10.11959/j.issn.2096-0271.2022026

摘要/Abstract

摘要：

在目前的中文文本情感分析研究中，基于语义规则和情感词典的方法通常需要人工设置情感阈值；而基于深度学习的方法由于未能运用语义规则和情感词典等语言知识，不能充分提取情感特征。针对这两种方法的缺点，提出了一种将语言知识和深度学习结合的文本情感分析方法。该方法首先根据语义规则提取文本中的关键情感片段，再根据情感词典从关键情感片段中抽取出情感更加明确的情感词来构建情感集合，然后利用深度学习模型分别从原始文本、关键情感片段、情感集合中抽取深层次特征，最后对提取的特征进行加权融合，并利用分类器实现情感极性的判断。实验结果表明，与未引入语言知识的深度学习模型相比，该方法的情感极性分类能力有明显提升。

关键词: 文本情感分析, 语言知识, 深度学习

Abstract:

At present, in the research of Chinese text emotion analysis, the method based on semantic rules and emotion dictionary usually needs to set the emotional threshold manually.However, the method based on deep learning can’t fully extract emotional features because it fails to use language knowledge such as semantic rules and emotional dictionary.As to shortcomings of two methods, a text emotion analysis method combining language knowledge and deep learning was proposed.Firstly, the key emotional segments in the text were extracted according to the semantic rules.Secondly, more explicit emotion words were extracted from the key emotional segments according to the emotional dictionary to construct the emotion set.Thirdly, the deep level features were extracted from the original text, key emotional segments and emotional set by using the deep learning model.Finally, the features were weighted and fused, and the classifier was used to judge the emotional polarity.The experimental results show that compared with the deep learning model without language knowledge, this method has significantly improved the ability of emotional polarity classification.

Key words: text sentiment analysis, language knowledge, deep learning

中图分类号:

TP391.1

徐康庭, 宋威. 结合语言知识和深度学习的中文文本情感分析方法[J]. 大数据, 2022, 8(3): 115-127.

Kangting XU, Wei Song. A Chinese text sentiment analysis method combining language knowledge and deep learning[J]. Big Data Research, 2022, 8(3): 115-127.

图/表 11

图1

图2

图3

图4

表1

表2

表3

表4

表5

表6

表7

参考文献 20

[1]	HATZIVASSILOGLOU V , MCKEOWN K R . Predicting the semantic orientation of adjectives[C]// Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and 8th Conference of the European Chapter of the Association for Computational Linguistics. Stroudsburg:Association for Computational Linguistics, 1997: 174-181.
[2]	WU J S , LU K , SU S Z ,et al. Chinese micro-blog sentiment analysis based on multiple sentiment dictionaries and semantic rule sets[J]. IEEE Access, 2019,7: 183924-183939.
[3]	赵妍妍, 秦兵, 石秋慧 ,等. 大规模情感词典的构建及其在情感分类中的应用[J]. 中文信息学报, 2017,31(2): 187-193.
	ZHAO Y Y , QIN B , SHI Q H ,et al. Large-scale sentiment lexicon collection and its application in sentiment classification[J]. Journal of Chinese Information Processing, 2017,31(2): 187-193.
[4]	XU G X , YU Z H , YAO H S ,et al. Chinese text sentiment analysis based on extended sentiment dictionary[J]. IEEE Access, 2019,7: 43749-43762.
[5]	KESHAVARZ H , ABADEH M S . ALGA:adaptive lexicon learning using genetic algorithm for sentiment analysis of microblogs[J]. Knowledge-Based Systems, 2017,122: 1-16.
[6]	李继东, 王移芝 . 基于扩展词典与语义规则的中文微博情感分析[J]. 计算机与现代化, 2018(2): 89-95.
	LI J D , WANG Y Z . Sentiment analysis of Chinese microblog based on expand-dictionary and semantic rule[J]. Computer and Modernization, 2018(2): 89-95.
[7]	ZHANG S X , WEI Z L , WANG Y ,et al. Sentiment analysis of Chinese microblog text based on extended sentiment dictionary[J]. Future Generation Computer Systems, 2018,81: 395-403.
[8]	PANG B , LEE L , VAITHYANATHAN S . Thumbs up?:sentiment classification using machine learning techniques[C]// Proceedings of the ACL-02 Conference on Empirical methods in Natural Language Processing. Stroudsburg:Association for Computational Linguistics, 2002: 79-86.
[9]	苏莹, 张勇, 胡珀 ,等. 基于朴素贝叶斯与潜在狄利克雷分布相结合的情感分析[J]. 计算机应用, 2016,36(6): 1613-1618.
	SU Y , ZHANG Y , HU P ,et al. Sentiment analysis research based on combination of naive Bayes and latent Dirichlet allocation[J]. Journal of Computer Applications, 2016,36(6): 1613-1618.
[10]	刘栋军, 王宇涵, 凌文芬 ,等. 基于脑机协同智能的情绪识别[J]. 智能科学与技术学报, 2021,3(1): 65-75.
	LIU D J , WANG Y H , LING W F ,et al. Emotion recognition based on brain and machine collaborative intelligence[J]. Chinese Journal of Intelligent Science and Technology, 2021,3(1): 65-75.
[11]	KIM Y , . Convolutional neural networks for sentence classification[C]// Proceedings of 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg:Association for Computational Linguistics, 2014: 1746-1751.
[12]	胡荣磊, 芮璐, 齐筱 ,等. 基于循环神经网络和注意力模型的文本情感分析[J]. 计算机应用研究, 2019,36(11): 3282-3285.
	HU R L , RUI L , QI X ,et al. Text sentiment analysis based on recurrent neural networks and attention model[J]. Application Research of Computers, 2019,36(11): 3282-3285.
[13]	李洋, 董红斌 . 基于CNN和BiLSTM网络特征融合的文本情感分析[J]. 计算机应用, 2018,38(11): 3075-3080.
	LI Y , DONG H B . Text sentiment analysis based on feature fusion of convolution neural network and bidirectional long short-term memory network[J]. Journal of Computer Applications, 2018,38(11): 3075-3080.
[14]	宋婷, 陈战伟, 杨海峰 . 基于分层注意力网络的方面情感分析[J]. 大数据, 2020,6(5): 82-91.
	SONG T , CHEN Z W , YANG H F . Aspect sentiment analysis based on a hierarchical attention network[J]. Big Data Research, 2020,6(5): 82-91.
[15]	谢润忠, 李烨 . 基于BERT和双通道注意力的文本情感分类模型[J]. 数据采集与处理, 2020,35(4): 642-652.
	XIE R Z , LI Y . Text sentiment classification model based on BERT and dual channel attention[J]. Journal of Data Acquisition and Processing, 2020,35(4): 642-652.
[16]	邱宁佳, 王晓霞, 王鹏 ,等. 融合语法规则的双通道中文情感模型分析[J]. 计算机应用, 2021,41(2): 318-323.
	QIU N J , WANG X X , WANG P ,et al. Analysis of double-channel Chinese sentiment model integrating grammar rules[J]. Journal of Computer Applications, 2021,41(2): 318-323.
[17]	CHUNG Y A , ZHU C G , ZENG M . SPLAT:speech-language joint pre-training for spoken language understanding[C]// Proceedings of 2021 Conference of the North American Chapter of Association for Computational Linguistics:Human Language Technologies. Stroudsburg:Association for Computational Linguistics, 2021: 1897-1907.
[18]	HOCHREITER S , SCHMIDHUBER J . Long short-term memory[J]. Neural Computation, 1997,9(8): 1735-1780.
[19]	GRAVES A . Supervised sequence labelling with recurrent neural networks[M]. Heidelberg: Springer, 2012.
[20]	CUI Z C , CHEN W L , CHEN Y X . Multiscale convolutional neural network for time series classification[J]. arXiv preprint, 2016,arXiv:1603.06995.

积极	消极
酒店环境、房间档次都很不错，服务水平专业。住了两天，感觉确实不错，二楼的自助烧烤给人宾至如归的感觉	房间装修陈旧，下水管堵塞，晚上折腾了两个多小时还是没有修好
非常好的酒店，四星级的标准，完全超值的享受，服务非常好	环境一般，住了之后让人感觉价格和服务不成正比
地点不错，邻近师范大学，到步行街走路15 min。房间很大很舒适	我肯定不会再住在这里了，太陈旧了，霉味太重，感觉不好

积极	消极
一如既往地好吃，希望可以开到其他城市	跟平时吃的味道一样，本来吃得很开心的，但是突然吃到一只虫，把我吐惨了，太恶心了，再也不吃了
老板很热情，店虽然小，但是味道非常好，很实惠	无良商家，排骨不是排骨就算了，肉还是臭的，臭得熏天，太恶心了
菜品不是特别多，但是这个价位已经很划算了	好咸，不好吃，非常不好吃

词典种类	数量/个	实例
情感词	27 466	索然寡味、一塌糊涂、杰出、宝贝
否定词	59	不必、徒然、毋庸、不
程度副词	219	极其、极度、非常、很

词性	数量/个	实例
转折	10	不料、但是、偏偏、岂知
递进	9	并且、更有甚者、况且、甚至
假设	12	那么、若、倘若、假如
因果	14	因此、所以、以致、以便
让步	10	尽管、虽然、纵使、即使

参数	值
BiLSTM隐层节点数	128
MCNN隐层节点数	256
DENSE隐层节点数	128
Kernel数量	3
Kernel_size	3、4、5
优化器	Adam
学习率	2×10^-5
Batch_size	32
词嵌入维度	768
激活函数	ReLU