Flexible and Creative Chinese Poetry Generation Using Neural Memory
39992Authors
nan
Abstract
AbstractIt has been shown that Chinese poems can be successfully generated by sequence-to-sequence neural models, particularly with the attention mechanism. A potential problem of this approach, however, is that neural models can only learn abstract rules, while poem generation is a highly creative process that involves not only rules but also innovations for which pure statistical models are not appropriate in principle. This work proposes a memory augmented neural model for Chinese poem generation, where the neural model and the augmented memory work together to balance the requirements of linguistic accordance and aesthetic innovation, leading to innovative generations that are still rule-compliant. In addition, it is found that the memory mechanism provides interesting flexibility that can be used to generate poems with different styles.
Keywords
Full Content
Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1364–1373 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1125 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1364–1373 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1125 Flexible and Creative Chinese Poetry Generation Using Neural Memory Jiyuan Zhang1,2, Yang Feng3,1,4, Dong Wang1∗, Yang Wang1, Andrew Abel6, Shiyue Zhang1,5, Andi Zhang1,5 1Center for Speech and Language Technologies(CSLT), RIIT, Tsinghua University, China 2Shool of Software & Microelectronics, Peking University, China 3Key Laboratory of Intelligent Information Processing,Institute of Computing Technology,CAS 4Huilan Limited, Beijing, China 5Beijing University of Posts and Telecommunications, China 6Xi’an Jiaotong-Liverpool University, China zhangjy ml@pku.edu.cn, wangdong99@mails.tsinghua.edu.cn Abstract It has been shown that Chinese po- ems can be successfully generated by sequence-to-sequence neural models, par- ticularly with the attention mechanism. A potential problem of this approach, how- ever, is that neural models can only learn abstract rules, while poem generation is a highly creative process that involves not only rules but also innovations for which pure statistical models are not appropri- ate in principle. This work proposes a memory-augmented neural model for Chi- nese poem generation, where the neural model and the augmented memory work together to balance the requirements of linguistic accordance and aesthetic inno- vation, leading to innovative generations that are still rule-compliant. In addition, it is found that the memory mechanis- m provides interesting flexibility that can be used to generate poems with different styles. 1 Introduction Classical Chinese poetry is a special cultural her- itage with over 2,000 years of history and is stil- l fascinating us today. Among the various gen- res, perhaps the most popular one is the quatrain, a special style with a strict structure (four lines with five or seven characters per line), a regulat- ed rhythmical form (the last characters in the sec- ond and fourth lines must follow the same rhyth- m), and a required tonal pattern (tones of charac- ters in some positions should satisfy a predefined regulation) (Wang, 2002). This genre flourished mostly in the Tang Dynasty, and so are often called 1Corresponding author: Dong Wang; RM 1-303, FIT BLDG, Tsinghua University, Beijing (100084), P.R. China. ‘Tang poems’. An example of a quatrain written by Wei Wang, a famous poet in the Tang Dynasty, is shown in Table 1. Due to the stringent restrictions in both rhyth- m and tone, it is not trivial to create a ful- ly rule-compliant quatrain. More importantly, besides such strict regulations, a good quatrain should also read fluently, hold a consistent theme, and express a unique affection. Therefore, poem generation is widely recognized as a very intelli- gent activity and can be performed only by knowl- edgeable people with a lot of training. Wi� Climbing the Paradise Mound •�¿Ø·§ (* Z Z P Z) As I was not in a good mood this evening round, °•���"(P P P Z P) I went by cart to …
Content truncated to 500 words
Labeled Text
AbstractIt has been shown that Chinese poems can be successfully generated by sequence-to-sequence neural models, particularly with the attention mechanism. A potential problem of this approach, however, is that neural models can only learn abstract rules, while poem generation is a highly creative process that involves not only rules but also innovations for which pure statistical models are not appropriate in principle. This work proposes a memory augmented neural model for Chinese poem generation, where the neural model and the augmented memory work together to balance the requirements of linguistic accordance and aesthetic innovation, leading to innovative generations that are still rule-compliant. In addition, it is found that the memory mechanism provides interesting flexibility that can be used to generate poems with different styles.
Publication Details
- Paper ID:
- 39992
- Year:
- 2017
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan