Chinese Poetry Generation with Recurrent Neural Networks

By Xingxing Zhang, Mirella Lapata. 2014

346 citations recorded by the source corpus.

We propose a model for Chinese poem generation based on recurrent neural networks which we argue is ideally suited to capturing poetic content and form. Our generator jointly performs content selection (“what to say”) and surface realization (“how to say”) by learning representations of individual characters, and their combinations into one or more lines as well as how these mutually reinforce and constrain each other. Poem lines are generated incrementally by taking into account the entire history of what has been generated so far rather than the limited horizon imposed by the previous line or lexical n-grams. Experimental results show that our model outperforms competitive Chinese poetry generation systems using both automatic and manual evaluation methods.

Subfields: Artificial Intelligence, Natural Language Processing, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Generative Models

Citations in the corpus

Neural Generation for Czech: Data and Baselines (2019, background): “Other works on neural non-English NLG solve in fact different tasks from ours: Chinese poetry generation (Zhang and Lapata, 2014; Yi et al., 2017; Wang et al., 2016), non-task-oriented response generation in chatbots (Xing et al.”
Hierarchical Neural Story Generation (2018, neutral): “, 2018) and poetry (Zhang and Lapata, 2014).”
SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient (2016, background): “However, in many other practical applications, such as poem generation (Zhang and Lapata 2014) and chatbot (Hingston 2009), a task speciﬁc loss may not be directly available to score a generated sequence accurately.”
Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems (2015, background): “…by sampling input tokens one by one from the output distribution of the RNN until a stop sign is gen-1 erated (Karpathy and Fei-Fei, 2014) or some constraint is satisﬁed (Zhang and Lapata, 2014), the network can produce a sequence of tokens which can be lexicalised 2 to form the required utterance.”
Neural Summarization by Extracting Sentences and Words (2016, methodology): “We ﬁrst obtain representation vectors at the sentence level using a single-layer convolutional neural network (CNN) with a max-over-time pooling operation (Kalchbrenner and Blun-som, 2013; Zhang and Lapata, 2014; Kim et al., 2016).”
Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking (2015, background): “Zhang and Lapata (2014) describes interesting work using RNNs to generate Chinese poetry.”
Maximum-Likelihood Augmented Discrete Generative Adversarial Networks (2017, methodology): “We use the dataset provided in (Zhang & Lapata, 2014), and split them in the standard way 1 .”
Summarizing Source Code using a Neural Attention Model (2016, neutral): “, 2015), and Chinese poetry generation (Zhang and Lapata, 2014).”
Long Text Generation via Adversarial Training with Leaked Information (2017, methodology): “Generating text that mimics human’s expression has been studied for poem generation (Zhang and Lapata 2014), image captioning (Vinyals et al. 2015), dialogue system (Li et al. 2017) machine translation (Yang et al. 2017).”
Recurrent Dropout without Memory Loss (2016, background): “…classiﬁcation (Wang et al., 2015; Irsoy and Cardie, 2014; Liu et al., 2015) and various tagging problems (Dyer et al., 2015) to language modelling (Kim et al., 2015; Zhang et al., 2015), text generation (Zhang and Lapata, 2014) and sequence-to-sequence prediction tasks (Sutskever et al., 2014).”
Survey of the State of the Art in Natural Language Generation: Core tasks, applications and evaluation (2017, methodology): “For instance, Zhang and Lapata (2014a) focus on poetry generation in Chinese using rnns, where verses are generated by predicting characters based on (a) the previous verses in the poem, represented by a convolutional sentence model (Kalchbrenner & Blunsom, 2013); and (b) the previous characters in…”
Adversarial Ranking for Language Generation (2017, methodology): “Following the evaluation protocol in [35, 37], we compute the BLEU-2 score and estimate the similarity between the human-written poem and the machine-created one.”
Hafez: an Interactive Poetry Generation System (2017, methodology): “…approach the problem by using grammatical and semantic templates (Oliveira, 2009, 2012) or treating the generation task as a translation/summarization task (Zhou et al., 2009; He et al., 2012; Yan et al., 2013; Zhang and Lapata, 2014; Yi et al., 2016; Wang et al., 2016; Ghazvininejad et al., 2016).”
FUDGE: Controlled Text Generation With Future Discriminators (2021, neutral): “been studied in different forms by many previous works (Zhang and Lapata, 2014; Wang et al., 2016; Ghazvininejad et al., 2016, 2017).”
Adversarial Generation of Natural Language (2017, background): “Zhang and Lapata (2014) present a dataset of Chinese poems that were used to evaluate adversarial training methods for natural language in (Yu et al., 2016) and (Che et al., 2017).”
Recent Advances in Recurrent Neural Networks (2017, methodology): “RNNs are also used to capture poetic style in works of literature and generate lyrics, for example Rap lyric generation [90]–[92].”
Natural Language Generation for Spoken Dialogue System using RNN Encoder-Decoder Networks (2017, background): “Zhang and Lapata (2014) proposed a generator using RNNs to create Chinese poetry.”
Topic-to-Essay Generation with Neural Networks (2018, background): “Chinese poetry generation [Zhang and Lapata, 2014; Wang et al. , 2016b] is to generate a kind of structural text, which contains some speciﬁc structural, rhythmical and tonal patterns.”