Towards Automatic Generation of Entertaining Dialogues in Chinese Crosstalks

TL;DR

This paper explores the first attempt to automatically generate entertaining Chinese crosstalk dialogues, proposing a humor-enhanced translation model that effectively produces supporting role responses based on leading utterances.

Contribution

It introduces a novel task of automatic Chinese crosstalk dialogue generation and proposes a humor-enhanced translation model to address this challenge.

Findings

Human evaluation confirms the model's effectiveness.

Automatic generation of entertaining dialogues is feasible.

The proposed model outperforms baseline methods.

Abstract

Crosstalk, also known by its Chinese name xiangsheng, is a traditional Chinese comedic performing art featuring jokes and funny dialogues, and one of China's most popular cultural elements. It is typically in the form of a dialogue between two performers for the purpose of bringing laughter to the audience, with one person acting as the leading comedian and the other as the supporting role. Though general dialogue generation has been widely explored in previous studies, it is unknown whether such entertaining dialogues can be automatically generated or not. In this paper, we for the first time investigate the possibility of automatic generation of entertaining dialogues in Chinese crosstalks. Given the utterance of the leading comedian in each dialogue, our task aims to generate the replying utterance of the supporting role. We propose a humor-enhanced translation model to address this task and human evaluation results demonstrate the efficacy of our proposed model. The feasibility of automatic entertaining dialogue generation is also verified.