Pun-GAN: Generative Adversarial Network for Pun Generation

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 3388–3393, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 3388 Pun-GAN: Generative Adversarial Network for Pun Generation Fuli Luo1∗, Shunyao Li1∗, Pengcheng Yang1, Lei li1, Baobao Chang1,2, Zhifang Sui1,2, Xu Sun1 1Key Lab of Computational Linguistics, Peking University 2Peng Cheng Laboratory, China {luofuli, lishunyao, yang pc, lilei nlp, chbb, szf, xusun}@pku.edu.cn Abstract In this paper, we focus on the task of generating a pun sentence given a pair of word senses. A major challenge for pun generation is the lack of large- scale pun corpus to guide the supervised learning. To remedy this, we propose an adversarial gen- erative network for pun generation (Pun-GAN), which does not require any pun corpus. It con- sists of a generator to produce pun sentences, and a discriminator to distinguish between the gener- ated pun sentences and the real sentences with spe- cific word senses. The output of the discrimina- tor is then used as a reward to train the generator via reinforcement learning, encouraging it to pro- duce pun sentences which can support two word senses simultaneously. Experiments show that the proposed Pun-GAN can generate sentences that are more ambiguous and diverse in both automatic and human evaluation.1 1 Introduction Generating creative and interesting text is a key step towards building an intelligent natural lan- guage generation system. A pun is a clever and amusing use of a word with two meanings (word senses), or of words with the same sound but dif- ferent meanings (Miller and Gurevych, 2015). In this paper, we focus on the former type of pun, i.e., homographic pun. For example, “I used to be a banker but I lost interest” is a pun sentence be- cause the pun word “interest” can be interpreted as either curiosity or profits. An intractable problem for pun generation is the lack of a large-scale pun corpus in which each pun sentence is labeled with two word senses. Early ∗Equal Contribution. 1The code is available at: https://github.com/ lishunyao97/Pun-GAN. !" !# $% &' Generator Discriminator word senses of the target pun word ( Fake ), + ~-./0/ Sample from non-pun data. Ambiguity reward Two senses of the target pun word ) is a sentence whose sense is + Figure 1: The proposed Pun-GAN framework. researches (Hong and Ong, 2009; Valitutti et al., 2013; Petrovic and Matthews, 2013) are mainly based on templates and rules, thus lacking creativ- ity and flexibility. Yu et al. (2018) is the first en- deavor to apply neural network to this task, which adopts a constrained neural language model (Mou et al., 2015) to guarantee that a pre-given word sense to appear in the generated sequence. How- ever, Yu et al. (2018) only integrates the generation probabilities of two word senses during the infer- ence decoding process, without detecting whether the generated sentences can support the two senses indeed during training. Promisingly, Word Sense …

Content truncated to 500 words