Generating Chinese Couplets and Quatrain Using a Statistical Approach
19381Authors
nan
Abstract
. We propose a novel statistical approach to automatically generate Chinese couplets and Chinese poetry. For Chinese couplets, the system takes as input the first sentence and generates as output an N-best list of second sentences using a phrase-based SMT model. A comprehensive evaluation using both human judgments and BLEU scores has been conducted and the results demonstrate that this approach is very successful. We then extended this approach to generate classic Chinese poetry using the quatrain as a case study. Given a few keywords describing a user's intention, a statistical model is used to generate the first sentence. Then a phrase-based SMT model is used to generate the other three quatrain sentences one by one. Evaluation using human judgment over individual lines as well as the quality of the generated poem as a whole demonstrates promising results. Keywords: Chinese couplets, Chinese classic poetry, automatic generation. ∗ This paper summarizes our work on automatic generation of Chinese couplets (Jiang & Zhou, 2008) and automatic generation of poetry (He et al., to appear), both were conducted in Natural Language Group at Microsoft Research Asia. Copyright 2009 by Ming Zhou, Long Jiang, and Jing He 1
Keywords
Full Content
Generating Chinese Couplets and Quatrain Using a Statistical Approach ∗∗∗∗ Ming Zhou a, Long Jiang a, and Jing He b a Microsft Research Asia, Sigma Centre, Haidian, Beijing, 100109, P.R. China {mingzhou, longj}@microsoft.com b Dept. of Computer Science, Tsinghua University hejing2929@gmail.com Abstract. We propose a novel statistical approach to automatically generate Chinese couplets and Chinese poetry. For Chinese couplets, the system takes as input the first sentence and generates as output an N-best list of second sentences using a phrase-based SMT model. A comprehensive evaluation using both human judgments and BLEU scores has been conducted and the results demonstrate that this approach is very successful. We then extended this approach to generate classic Chinese poetry using the quatrain as a case study. Given a few keywords describing a user's intention, a statistical model is used to generate the first sentence. Then a phrase-based SMT model is used to generate the other three quatrain sentences one by one. Evaluation using human judgment over individual lines as well as the quality of the generated poem as a whole demonstrates promising results. Keywords: Chinese couplets, Chinese classic poetry, automatic generation. ∗ This paper summarizes our work on automatic generation of Chinese couplets (Jiang & Zhou, 2008) and automatic generation of poetry (He et al., to appear), both were conducted in Natural Language Group at Microsoft Research Asia. Copyright 2009 by Ming Zhou, Long Jiang, and Jing He 1 Introduction This paper presents a novel approach to generate Chinese couplets and classic Chinese poetry with a Statistical Machine Translation (SMT) model. Chinese antithetical couplets, called “duìlián”, form a special type of poetry composed of two sentences. The two sentences making up a couplet are called the “first sentence” (FS) and the “second sentence” (SS) respectively. The quatrain and regulated verse are the two most important forms of classic Chinese poetry. A quatrain is composed of 4 lines and regulated verse is composed of 8 lines with additional rules governing structure. Each line of the quatrain or regulated verse is composed of 5 Chinese characters (五言) or 7 characters (七言). For Chinese couplets, the general process of generating a SS given a FS is like this: for each word in the FS, find some words that can be used as the counterparts in the SS; then from the word lattice, select one word at each position in the SS so that the selected words form a fluent sentence satisfying the constraints of Chinese couplets. This process is similar to translating a source language sentence into a target language sentence without word insertion, deletion and reordering, but the target sentence should satisfy some linguistic constraints. Based on this observation, we propose a multi-phase statistical machine translation approach to generate the SS. First, a phrase-based statistical machine translation (SMT) model is applied to generate an 43 23rd Pacific Asia Conference on Language, Information and Computation, pages 43–52 N-best list of SS candidates. Then, a set of filters based on linguistic constraints for Chinese couplets is used to …
Content truncated to 500 words
Labeled Text
. We propose a novel statistical approach to automatically generate Chinese couplets and Chinese poetry. For Chinese couplets, the system takes as input the first sentence and generates as output an N-best list of second sentences using a phrase-based SMT model. A comprehensive evaluation using both human judgments and BLEU scores has been conducted and the results demonstrate that this approach is very successful. We then extended this approach to generate classic Chinese poetry using the quatrain as a case study. Given a few keywords describing a user's intention, a statistical model is used to generate the first sentence. Then a phrase-based SMT model is used to generate the other three quatrain sentences one by one. Evaluation using human judgment over individual lines as well as the quality of the generated poem as a whole demonstrates promising results. Keywords: Chinese couplets, Chinese classic poetry, automatic generation. ∗ This paper summarizes our work on automatic generation of Chinese couplets (Jiang & Zhou, 2008) and automatic generation of poetry (He et al., to appear), both were conducted in Natural Language Group at Microsoft Research Asia. Copyright 2009 by Ming Zhou, Long Jiang, and Jing He 1
Publication Details
- Paper ID:
- 19381
- Year:
- 2009
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan