基于多样性情绪的积极导向自然过渡决策模型

doi:10.3778/j.issn.1002-8331.2406-0001

摘要/Abstract

摘要： 在对话系统中情绪生成任务旨在生成待回复话语中的情绪类别。针对现有情绪生成方法中个性化情绪积极自然过渡机制的缺失，造成情绪状态生成连贯性问题，提出积极导向自然过渡决策模型（positive orientation and natural transition decision model，PONTD）用于进行多样化情绪预测后情绪决策，以解决情绪生成连贯性问题。PONTD包含三个单元：一是多样化情绪预测单元，设计多组不同结构和参数的模型，并采用共同注意力融合方法，实现对情绪状态的多样化预测；二是情绪正向引导单元，利用预测的多样性情绪状态，赋予对话人格，引导情绪根据情绪极性进行转化，并利用SetRank模型将情绪进行重排，以实现个性化积极的情绪导向；三是个性化自然过渡单元，通过对多样性情绪状态进行关联感知，根据人格PAD情绪空间映射规则进行情绪矫正，并利用对话历史情绪序列控制时机，实现情绪的自然平滑过渡。实验结果表明，PONTD能有效提升情绪表达的连贯性，提高了对话系统情绪智能。

关键词: 对话系统, 情绪生成, 个性化情绪, 情绪决策, 对话人格

Abstract: In the task of emotion generation in dialogue systems, the objective is to generate the emotional category in the response sentences. Addressing the lack of a personalized mechanism for a positive and natural transition of emotions in existing emotion generation methods, which leads to coherence issues in the generation of emotional states, this paper proposes the positive orientation and natural transition decision model （PONTD） for diversified emotional prediction and decision-making to solve the problem of emotional generation coherence. PONTD consists of three components： first, a diversified emotional prediction unit, where multiple models with different structures and parameters are designed and a common attention fusion method is employed to predict various emotional states; second, a positive emotional orientation unit, which utilizes the predicted diverse emotional states to endow the dialogue with personality, guides the transition of emotions based on their polarity, and employs the SetRank model to reorder emotions, achieving personalized positive emotional guidance; and third, a personalized natural transition unit, which correlates the diverse emotional states through associative perception, corrects emotions according to personality PAD emotion space mapping rules, and uses the dialogue history emotion sequence to control timing, realizing a natural and smooth transition of emotions. Experimental results show that PONTD can effectively enhance the coherence of emotional expression and improve the emotional inte-lligence of dialogue systems.

Key words: dialogue system, emotion generation, personalized emotion, emotion decision, dialogue personality

马志强, 吕凯, 周钰童, 刘佳, 叶浩然, 刘义兴, 王奎波. 基于多样性情绪的积极导向自然过渡决策模型[J]. 计算机工程与应用, 2025, 61(17): 241-250.

MA Zhiqiang, LYU Kai, ZHOU Yutong, LIU Jia, YE Haoran, LIU Yixing, WANG Kuibo. Positive Orientation and Natural Transition Decision Model Based on Diverse Emotions[J]. Computer Engineering and Applications, 2025, 61(17): 241-250.

参考文献

[1] PICARD R W. Affective computing[M]. London: MIT Press, 2000.
[2] OUYANG L, WU J, JIANG X, et al. Training language models to follow instructions with human feedbackC]//Advances in Neural Information Processing Systems, 2022: 27730-27744.
[3] ZHAO W, LI Z, WANG S, et al. Both matter: enhancing the emotional intelligence of large language models without compromising the general intelligence[J]. arXiv:2402.10073, 2024.
[4] MAYER J D, SALOVEY P, CARUSO D R, et al. Emotional intelligence as a standard intelligence[J]. Emotion, 2001, 1(3): 232-242.
[5] GHOSH S, CHOLLET M, LAKSANA E, et al. Affect-LM: a neural language model for customizable affective text generation[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics. Stroudsburg: ACL, 2017: 634-642.
[6] ZHOU H, HUANG M L, ZHANG T Y, et al. Emotional chatting machine: emotional conversation generation with internal and external memoryC]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018: 730-739.
[7] MA Z Q, JIA W C, ZHOU Y T, et al. Personality enhanced emotion generation modeling for dialogue systems[J]. Cognitive Computation, 2024, 16(1): 293-304.
[8] 周钰童, 马志强, 许璧麒, 等. 基于深度学习的对话情绪生成研究综述[J]. 计算机工程与应用, 2024, 60(7): 13-25.
ZHOU Y T, MA Z Q, XU B Q, et al. Survey of deep learning-based on emotion generation in conversation[J]. Computer Engineering and Applications, 2024, 60(7): 13-25.
[9] WEI W, LIU J Y, MAO X L, et al. Emotion-aware chat mac-hine: automatic emotional response generation for human-like emotional interaction[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management. New York: ACM, 2019: 1401-1410.
[10] GU X S, XU W R, LI S. Towards automated emotional conversation generation with implicit and explicit affective strategy[C]//Proceedings of the 2019 International Symposium on Signal Processing Systems. New York: ACM, 2019: 125-130.
[11] GONG M M, SONG H, ZHOU H R, et al. Enhancing resp-onse relevance and emotional consistency for dialogue resp-onse generation[C]//Proceedings of the 17th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAI-NLP). Piscataway: IEEE, 2022: 1-6.
[12] ZHANG R, WANG Z Y, HUANG Z H, et al. Predicting emotion reactions for human computer conversation: a variational approach[J]. IEEE Transactions on Human-Machine Systems, 2021, 51(4): 279-287.
[13] WEN Z Y, CAO J N, YANG R S, et al. Automatically select emotion for response via personality-affected emotion transition[C]//Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Stroudsburg: ACL, 2021: 5010-5020.
[14] 马志强, 王春喻, 贾文超, 等. 情感导向对话回复生成模型[J]. 中文信息学报, 2023, 37(8): 104-114.
MA Z Q, WANG C Y, JIA W C, et al. Emotion guided dialogue response generation model[J]. Journal of Chinese Information Processing, 2023, 37(8): 104-114.
[15] LUBIS N, SAKTI S, YOSHINO K, et al. Eliciting positive emotion through affect-sensitive dialogue response generation: a neural network approach[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018: 5293-5300.
[16] 李大宇. 多轮文本对话中的情绪预测方法研究[D]. 太原: 山西大学, 2021.
LI D Y. Research on emotion prediction in multi-turn textual conversations[D]. Taiyuan: Shanxi University, 2021.
[17] YANG L, SHEN Y, MAO Y, et al. Hybrid curriculum learning for emotion recognition in conversation[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2022: 11595-11603.
[18] LIU Y C, ZHAO J M, HU J W, et al. DialogueEIN: emotion interaction network for dialogue affective analysis[C]//Proceedings of the International Conference on Computational Linguistics, 2022 : 684-693.
[19] ZHAO W X, ZHAO Y Y, WANG S L, et al. TransESC: smoothing emotional support conversation via turn-level state transition[C]//Findings of the Association for Computational Linguistics: ACL 2023. Stroudsburg: ACL, 2023: 6725-6739.
[20] KIM W, AHN Y, KIM D, et al. Emp-RFT: empathetic resp-onse generation via recognizing feature transitions between utterances[C]//Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACL, 2022: 4118-4128.
[21] LIU Y, OTT M, GOYAL N, et al. RoBERTa: a robustly optimized BERT pretraining approach[J]. arXiv:1907.11692, 2019.
[22] SHARMA A, MINER A, ATKINS D, et al. A computational approach to understanding empathy expressed in text-based mental health support[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2020: 5263-5276.
[23] ZOU Y C, LIU Z H, HU X W, et al. Thinking clearly, talking fast: concept-guided non-autoregressive generation for open-domain dialogue systems[C]//Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2021: 2215-2226.
[24] LI B B, FEI H, SU F F, et al. Integrating discourse features and response assessment for advancing empathetic dialogue[J]. Information Processing & Management, 2024, 61(5): 103803.
[25] MAJUMDER N, PORIA S, HAZARIKA D, et al. Dialogue-RNN: an attentive RNN for emotion detection in convers-ations[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2019: 6818-6825.
[26] GHOSAL D, MAJUMDER N, PORIA S, et al. Dialogue-GCN: a graph convolutional neural network for emotion recognition in conversation[C]//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing. Stroudsburg: ACL, 2019: 154-164.
[27] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems, 2017.
[28] WILDER J W. New concepts in technical trading systems[M]. Greensboro, NC: Trend Research, 1978.
[29] 黄宏程, 李净, 胡敏, 等. 基于强化学习的机器人认知情感交互模型[J]. 电子与信息学报, 2021, 43(6): 1781-1788.
HUANG H C, LI J, HU M, et al. Cognitive emotional interaction model of robot based on reinforcement learning[J]. Journal of Electronics & Information Technology, 2021, 43(6): 1781-1788.
[30] 黄宏程, 刘宁, 胡敏, 等. 基于博弈的机器人认知情感交互模型[J]. 电子与信息学报, 2019, 41(10): 2471-2478.
HUANG H C, LIU N, HU M, et al. Cognitive emotion interaction model of robot based on game theory[J]. Journal of Electronics & Information Technology, 2019, 41(10): 2471-2478.