School of Computer & Information Engineering, Jiangxi Normal University, Nanchang 330022, China
Show less
文章历史+
Received
Accepted
Published
2024-10-31
2025-03-08
Issue Date
2025-10-09
PDF (2255K)
摘要
细粒度情绪识别模型采用几十种情绪类别对人类情绪进行建模,能够比传统模型更准确地捕捉人们表达的细微情绪。然而,现有的情绪预测模型并未充分考虑到数量众多的细粒度情绪之间存在的复杂相关性。针对这一问题,本文提出了一种融合VAD(Valence-Arousal-Dominance,效价-唤醒-支配)知识的情感分布增强细粒度情绪识别方法(VAD Emotion Distribution Augmented Fine-grained Emotion Recognition,EDAER)。EDAER模型采用情感分布建模VAD空间中情绪的相关性,结合文本语义信息与心理学先验知识进行细粒度情绪识别。EDAER模型包括语义信息模块、情感分布信息模块和融合预测模块。语义信息模块通过预训练的BERT(Bidirectional Encoder Representation from Transformers)模型提取文本的语义特征;情感分布信息模块基于VAD距离度量情绪间的相似度,为情感词生成情感分布;融合预测模块则利用注意力机制将文本语义信息与情感分布信息整合,并实现情绪预测。在GoEmotions数据集上的实验结果表明,EDAER模型的宏平均F1值达到51.75%,优于使用情感词典作为外部知识的KEA(Knowledge-Embedded Attention)模型和采用情绪层级关系作为外部知识的HGCN-EC(Hierarchy Graph Convolution Networks based Emotion Recognition)模型。特别是在三个样本量较少的情绪类别上,EDAER的F1值显著高于其他模型。实验结果验证了通过情感分布建模VAD空间中情绪的相关性,可以有效学习罕见情绪的相关知识,从而提升模型对细粒度情绪的识别能力。
Abstract
Fine-grained emotion recognition models, which employ dozens of emotion categories to model human emotions, are capable of capturing subtle emotional expressions more accurately than traditional models. However, existing emotion prediction models have not fully considered the complex correlations that exist among the numerous fine-grained emotions. To address this issue, this paper proposes an VAD (Valence-Arousal-Dominance) Emotion Distribution Augmented Fine-grained Emotion Recognition (EDAER). EDAER models the emotional correlations in the VAD space using emotion distributions and combines textual semantic information with psychological priors for fine-grained emotion recognition. The EDAER model consists of three modules: a semantic information module, an emotion distribution information module, and a fusion prediction module. The semantic information module extracts textual semantic features using a pre-trained BERT (Bidirectional Encoder Representation from Transformers) model; the emotion distribution information module generates emotion distributions for emotion words based on VAD distance metrics to measure the similarity between emotions; and the fusion prediction module integrates textual semantic features and emotion distribution information through an attention mechanism to predict emotions. Experimental results on the GoEmotions dataset demonstrate that the macro-average F1 score of the EDAER model reaches 51.75%, outperforming both the KEA (Knowledge-Embedded Attention) model, which uses emotion lexicons as external knowledge, and the HGCN-EC (Hierarchy Graph Convolution Networks based Emotion Recognition) model, which utilizes hierarchical emotion relationships as external knowledge. Notably, for three emotion categories with fewer samples, EDAER significantly outperforms other models in terms of F1 score. These results validate that modeling emotional correlations in the VAD space through emotion distributions can effectively capture knowledge related to rare emotions, thus improving the model's ability to recognize fine-grained emotions.
虽然细粒度情绪模型提供了更丰富的情绪表达,但大量增加的情绪类别以及它们之间的相互关联和模糊性,给细粒度情绪识别模型带来了挑战。为了解决这些挑战,已有的细粒度情绪识别研究通过引入心理学情绪知识来提升模型性能。例如,Bruyne等[17]将BERT的上下文表示与词典分数结合起来,并通过Bi-LSTM进行分类;Dhar等[12]使用VADER(Valence Aware Dictionary and sEntiment Reasoner)词典对Tweet文本进行情绪分类;Zhong等[13]基于NRC-VAD(National Research Council Canada - Valence, Arousal, Dominance)词典中情感词的VAD分数作为外部知识,提升模型细粒度情绪识别性能。上述这些方法只采用情感词的VAD三维得分,没有充分考虑VAD空间中情绪间的相关性。
为解决上述问题,本文提出融合VAD知识的情感分布增强细粒度情绪识别模型(VAD Emotion Distribution Augmented BERT for Fine-grained Emotion Recognition,EDAER),通过情感分布的方式量化每个情感词在VAD情绪模型空间中各情绪类别上的表达强度,如图3所示。EDAER模型构架如图4所示,包括语义信息模块、情感分布信息模块和融合预测模块。
融合预测模块的作用是将语义信息和情感分布信息结合,通过注意力机制和拼接操作实现信息融合并进行预测。如图4所示,首先,通过情感分布信息模块提取句子中的情感词,并将其转换为情感分布序列;其次,通过全连接层将情感分布序列转化为情感知识编码,;然后,利用自注意力机制,其中作为键(Key, K ),作为查询(Query, Q ), K 和 Q 做矩阵乘法经过Softmax得到注意力分数,将再作为值(Value, V )与做矩阵乘法;
● TextRCNN(Recurrent Convolutional Neural Networks for Text Classification)模型: TextRCNN模型是结合了卷积神经网络和时间序列神经网络来构建的文本分类模型。单层双向RNN用于学习上下文信息,并通过最大池化层进行特征选择以及全连接分类器。
CARDONEB, DI MARTINOF, SENATORES. Improving the Emotion-based Classification by Exploiting the Fuzzy Entropy in FCM Clustering[J]. Int J Intell Syst, 2021, 36(11): 6944-6967. DOI: 10.1002/int.22575 .
HUH J, YANGY Y, YIY, et al. An Emotion Classification Method Based on Fine-grained Information Perception BERT-EEP[J]. Comput Eng Sci, 2023, 45(4): 751-760. DOI: 10.3969/j.issn.1007-130X.2023.04.023 .
[4]
KHANPOURH, CARAGEAC. Fine-grained emotion detection in health-related online posts[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Brussels: ACL, 2018: 1160-1166.DOI:10.18653/v1/D18-1147 .
[5]
EKMANP. An Argument for Basic Emotions[J]. Cogn Emot, 1992, 6(3/4): 169-200. DOI: 10.1080/02699939208411068 .
[6]
PLUTCHIKR. The Nature of Emotions[J]. Am Sci, 2001, 89(4): 344. DOI: 10.1511/2001.4.344 .
ZHANGH T, WANGD, XUH L, et al. Sentiment Classification of Micro-blog Public Opinion Based on Convolution Neural Network[J]. J China Soc Sci Tech Inf, 2018, 37(7): 695-702. DOI: 10.3772/j.issn.1000-0135.2018.07.005 .
[9]
LEIZ Y, YANGY J, YANGM, et al. A Multi-sentiment-resource Enhanced Attention Network for Sentiment Classification[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). Stroudsburg, PA, USA: ACL, 2018: 758-763. DOI: 10.18653/v1/p18-2120 .
[10]
KELTNERD, LERNERJ S. Handbook of Social Psychology[M]. New Jersey: Wiley, 2010: 317-352. DOI: 10.1007/978-94-007-6772-0 .
[11]
COWENA S, ELFENBEINH A, LAUKKAP, et al. Mapping 24 Emotions Conveyed by Brief Human Vocalization[J]. Am Psychol, 2019, 74(6): 698-712. DOI: 10.1037/amp0000399 .
[12]
COWENA S, KELTNERD. What the Face Displays: Mapping 28 Emotions Conveyed by Naturalistic Expression[J]. Am Psychol, 2020, 75(3): 349-364. DOI: 10.1037/amp0000488 .
[13]
DEMSZKYD, MOVSHOVITZ-ATTIASD, KO J, et al. GoEmotions: A Dataset of Fine-grained Emotions[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA, USA: ACL, 2020: 4040-4054. DOI: 10.18653/v1/2020.acl-main.372 .
[14]
DHARS, BOSEI. Emotions in Twitter Communication and Stock Prices of Firms: The Impact of Covid-19 Pandemic[J]. Decision, 2020, 47(4): 385-399. DOI: 10.1007/S40622-020-00264-4 .
[15]
ZHONGP X, WANGD, MIAOC Y. Knowledge-enriched Transformer for Emotion Detection in Textual Conversations[C]//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). Stroudsburg, PA, USA: ACL, 2019: 165-176. DOI: 10.18653/v1/d19-1016 .
[16]
RUSSELLJ A, MEHRABIANA. Evidence for a Three-factor Theory of Emotions[J]. J Res Pers, 1977, 11(3): 273-294. DOI: 10.1016/0092-6566(77)90037-X .
[17]
SURESHV, ONGD C. Using Knowledge-embedded Attention to Augment Pre-trained Language Models for Fine-grained Emotion Recognition[C]//2021 9th International Conference on Affective Computing and Intelligent Interaction (ACII). New York: IEEE, 2021: 1-8. DOI: 10.1109/acii52823.2021.9597390 .
LAIJ S, WANZ Y, ZENGX Q. The Image Emotion Distribution Learning Based on Emotion Wheel and Multi-task Convolutional Neural Network[J]. J Jiangxi Norm Univ Nat Sci Ed, 2022, 46(4): 363-371. DOI: 10.16357/j.cnki.issn1000-5862.2022.04.06 .
[20]
DE BRUYNEL, ATANASOVAP, AUGENSTEINI. Joint Emotion Label Space Modeling for Affect Lexica[J]. Comput Speech Lang, 2022, 71: 101257. DOI: 10.1016/j.csl.2021.101257 .
[21]
ALSWAIDANN, BACHIR MENAI MEL. A Survey of State-of-the-art Approaches for Emotion Recognition in Text[J]. Knowl Inf Syst, 2020, 62(8): 2937-2987. DOI: 10.1007/s10115-020-01449-0 .
[22]
HON H, YANGH J, KIMS H, et al. Multimodal Approach of Speech Emotion Recognition Using Multi-level Multi-head Fusion Attention-based Recurrent Neural Network[J]. IEEE Access, 2020, 8: 61672-61686. DOI: 10.1109/access.2020.2984368 .
[23]
DEVLINJ, CHANGM W, LEEK, et al. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding[C]//Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics. Minneapolis, Minnesota: Association for Computational Linguistics. Stroudsburg, PA, USA: ACL, 2019: 4171-4186. DOI: 10.18653/v1/N19-1423 .
[24]
LIQ, WEIC, DANGJ N, et al. Tracking and Analyzing Public Emotion Evolutions During COVID-19: A Case Study from the Event-driven Perspective on Microblogs[J]. Int J Environ Res Public Health, 2020, 17(18): 6888. DOI: 10.3390/ijerph17186888 .
[25]
AKHTARS, GHOSALD, EKBALA, et al.All-in-One: Emotion, Sentiment and Intensity Prediction Using a Multi-task Ensemble Framework[J]. IEEE Trans Affect Comput, 2019, 13(1): 285-299. DOI: 10.1109/TAFFC.2019.2926724 .
[26]
WANGR Z, TANGD Y, DUANN, et al. K-Adapter: Infusing Knowledge into Pre-trained Models with Adapters[C]//Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Stroudsburg, PA, USA: ACL, 2021: 1405-1418. DOI: 10.18653/v1/2021.findings-acl.121 .
[27]
ABASA R, ELHENAWYI, ZIDANM, et al. BERT-CNN: a Deep Learning Model for Detecting Emotions from Text[J]. Comput Mater Continua, 2022, 71(2): 2943-2961. DOI: 10.32604/cmc.2022.021671 .
[28]
KUMARP, RAMANB. A BERT Based Dual-channel Explainable Text Emotion Recognition System[J]. Neural Netw, 2022, 150: 392-407. DOI: 10.1016/j.neunet.2022.03.017 .