摘 要：在文本生成领域中，古典诗词的自动生成研究是一个非常具有挑战性的任务。模型采用生成式对抗网络结构，将判别器内部的特征信息引入到生成器中，并作为生成器的指导信息，用于古诗词的生成。在判别器和生成器中分别加入情感判别器、情感门控机制，来将情感信息融入到古诗词中。实验结果表明，文章提出的模型所创作的诗歌不仅更加贴合古诗词韵律，且情感表达更加流畅，在不影响诗歌质量的前提下，丰富了古诗词的情感内容。
关键词：生成式对抗网络；古诗词生成；情感表达
DOI:10.19850/j.cnki.2096-4706.2022.21.002
中图分类号：TP391 文献标识码：A 文章编号：2096-4706（2022）21-0007-05
Emotional Ancient Poetry Generation Model Based on LeakGAN
ZHANG Yanliang
(School of Information Engineering, Dalian Ocean University, Dalian 116023, China)
Abstract: In the field of text generation, the automatic generation research of classical poetry is a very challenging task. The model adopts the structure of Generative Adversarial Network, introduces the characteristic information inside the discriminator into the generator and uses it as the guidance information of the generator to generate ancient poetry. Emotional discriminator and emotional gating mechanism are added to the discriminator and generator respectively, so as to integrate emotional information into ancient poetry. The experimental results show that the poetry created by the model proposed in this paper not only fits the rhythm of ancient poetry more closely, but also has more emotion expression. Under the premise of not affecting the quality of poetry, it enriches the emotional content of ancient poetry.
Keywords: Generative Adversarial Network; ancient poetry generation; emotion expression
参考文献：
[1] HARTMAN C. Virtual Muse ：Experiments in Computer Poctry [M]. Middletown：Wesleyan University Press，1996：95-96.
[2] AGNAR A，ENRIC P. Case-based reasoning：foundational issues. methodologicalvariations，and system approaches [J]. AICommunications，1994，7（1）：39-59.
[3] DÍAZAGUDO B，GERVÁS P，GONZÁLEZCALERO P A. Poetry generation in COLIBRI [J]. Ad Hoc Networks，2002，7（5）： 973-986.
[4] KEMPE V，LEVY R，GRACI C. Neural networks as fitness evaluators in genetic algorithms：Simulating human creativity [J/OL]// Proceedings of the Annual Meeting of the Cognitive Science Society. 2001，23(23)：[2022-06-09]. https://escholarship.org/uc/item/2tf5m4v5.
[5] JIANG L，ZHOU M. Generating Chinese couplets using a statisticalMT approach [C].Proceedings of the 22nd International Conference on Computational Linguistics Volume l.Stroudsburg： Association for ComputationalLinguistics，2008：377-384.
[6] GOODFELLOW I，POUGET-ABADIE J，MIRZA M， et al. Generative adversarial nets [C]//Advances in Neural Information Processing Systems.NeurIPS，2014：2672-2680.
[7] 马露育 . 基于改进的序列生成对抗网络的诗歌生成问题研究 [D]. 上海 : 上海师范大学，2020.
[8] 段明君 . 基于生成对抗网络的中文文本生成 [D]. 成都 :电子科技大学，2021.
[9] TIAN H S，YANG K X，LIU D H，et al. AnchiBERT：A Pre-Trained Model for Ancient Chinese Language Understanding and Generation [C]//2021 International Joint Conference on Neural Networks （IJCNN）. Shenzhen：IEEE，2021：1-8.
[10] BAO C，HUANG L. Chinese Traditional Poetry Generating System Based on Deep Learning [J/OL].arXiv：2110.12335 [cs.CL]. （2021-10-24）.https://arxiv.org/abs/2110.12335.
[11] GUO J X，LU S D，CAI H，et al. Long text generation via adversarial training with leaked information [J/OL].arXiv:1709.08624 [cs.CL].[2022-08-10].https://arxiv.org/abs/1709.08624.
[12] CHEN H，YI X，SUN M，et al. Sentiment-Controllable Chinese Poetry Generation [C]//Proceedings of the TwentyEighth International Joint Conference on Artificial Intelligence （IJCAI-19）.2019：4925-4931.
[13] YAN Y，SHEN G，ZHANG S，et al. Sequence generative adversarial nets with a conditional discriminator [J].Neurocomputing， 2021，429：69-76.
[14] 韩晓明 . 基于生成对抗网络的带关键词约束的情感文本生成 [D]. 厦门 : 厦门大学 ,2019.
[15] 庞栓栓 . 基于 LeakGAN 的诱饵文档生成研究与实现 [D].北京 : 北京交通大学，2019.
[16] 孙可佳，李启南 . 基于改进生成对抗网络的诗歌生成 [J].兰州交通大学学报，2020，39（2）：64-70.
[17] 廖荣凡，沈希忠，刘爽 . 更具有感情色彩的诗歌生成模型 [J]. 计算机系统应用，2020，29（5）：46-51.
作者简介：张彦亮 (1994—)，男，汉族，河北唐山人，硕士研究生在读，研究方向：古诗词文本生成。