Generating Major Types of Chinese Classical Poetry in a Uniformed Framework

14381

Authors

nan

Abstract

AbstractPoetry generation is an interesting research topic in the field of text generation. As one of the most valuable literary and cultural heritages of China, Chinese classical poetry is very familiar and loved by Chinese people from generation to generation. It has many particular characteristics in its language structure, ranging from form, sound to meaning, thus is regarded as an ideal testing task for text generation. In this paper, we propose a GPT-2 based uniformed framework for generating major types of Chinese classical poems. We define a unified format for formulating all types of training samples by integrating detailed form information, then present a simple form- stressed weighting method in GPT-2 to strengthen the control to the form of the generated poems, with special emphasis on those forms with longer body length. Preliminary experimental results show this enhanced model can generate Chinese classical poems of major types with high quality in both form and content, validating the effectiveness of the proposed strategy. The model has been incorporated into Jiuge, the most influential Chinese classical poetry generation system developed by Tsinghua University.

Keywords

nan

Full Content

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4658–4663 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4658 Generating Major Types of Chinese Classical Poetry in a Uniformed Framework Jinyi Hu, Maosong Sun∗ Department of Computer Science and Technology, Tsinghua University, Beijing, China Institute for Artificial Intelligence, Tsinghua University, Beijing, China State Key Lab on Intelligent Technology and Systems, Tsinghua University, Beijing, China hujy17@mails.tsinghua.edu.cn, sms@tsinghua.edu.cn Abstract Poetry generation is an interesting research topic in the field of text generation. As one of the most valuable literary and cultural heritages of China, Chinese classical poetry is very familiar and loved by Chinese people from generation to generation. It has many particular characteristics in its language structure, ranging from form, sound to meaning, thus is regarded as an ideal testing task for text generation. In this paper, we propose a GPT-2 based uniformed framework for generating major types of Chinese classical poems. We define a unified format for formulating all types of training samples by integrating form information, then present a simple form-stressed weighting method in GPT-2 to strengthen the control to the form of generated poems, with special emphasis on those forms with longer body length. Preliminary experimental results show this enhanced model can generate Chinese classical poems of major types with high quality in both form and content, validating the effectiveness of the proposed strategy. The model has been incorporated into Jiuge, the most influential Chinese classical poetry generation system developed by Tsinghua University (Guo et al., 2019). Keywords: poetry generation, GPT-2, form control 1. Introduction Chinese poetry is a rich treasure in Chinese traditional cul- ture. For thousands of years, poetry is always considered as the crystallization of human wisdom and erudition by Chi- nese people and deeply influences the Chinese history from the mental and cultural perspective. In general, a Chinese classical poem is a perfect combi- nation of three aspects, i.e., form, sound, and meaning. Firstly, it must strictly obey a particular form which spec- ifies the number of lines (i.e., sentences) in the poem and the number of characters in each line. Secondly, it must strictly obey a particular sound pattern which specifies the sound requirement for each character in every position of the poem. Lastly, it must be meaningful, i.e., with gram- matical and semantic well-formedness for each line and, with thematic coherence and integrity throughout the poem. These three points form the universal principles for human poets to create Chinese classical poems. Chinese Classical poetry can be classified into two primary categories, SHI and CI. According to the statistical data from CCPC1.0, a Chinese Classical Poetry Corpus consist- ing of 834,902 poems in total (We believe it is almost a full collection of Chinese Classical poems). 92.87% poems in CCPC1.0 fall into the category of SHI and 7.13% fall into the category of CI. SHI and CI can be further divided into many different types in terms of their forms. We briefly introduce the …

Content truncated to 500 words

Labeled Text

AbstractPoetry generation is an interesting research topic in the field of text generation. As one of the most valuable literary and cultural heritages of China, Chinese classical poetry is very familiar and loved by Chinese people from generation to generation. It has many particular characteristics in its language structure, ranging from form, sound to meaning, thus is regarded as an ideal testing task for text generation. In this paper, we propose a GPT-2 based uniformed framework for generating major types of Chinese classical poems. We define a unified format for formulating all types of training samples by integrating detailed form information, then present a simple form- stressed weighting method in GPT-2 to strengthen the control to the form of the generated poems, with special emphasis on those forms with longer body length. Preliminary experimental results show this enhanced model can generate Chinese classical poems of major types with high quality in both form and content, validating the effectiveness of the proposed strategy. The model has been incorporated into Jiuge, the most influential Chinese classical poetry generation system developed by Tsinghua University.

Publication Details

Paper ID:
14381
Year:
2020
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan