ConFiguRe: Exploring Discourse-level Chinese Figures of Speech

Proceedings of the 29th International Conference on Computational Linguistics, pages 3374–3385 October 12–17, 2022. 3374 ConFiguRe: Exploring Discourse-level Chinese Figures of Speech Dawei Zhu Qiusi Zhan Zhejian Zhou Yifan Song Jiebin Zhang Sujian Li∗ School of Computer Science, Peking University Key Laboratory of Computational Linguistics, MOE, Peking University† Abstract Figures of speech, such as metaphor and irony, are ubiquitous in literature works and collo- quial conversations. This poses great challenge for natural language understanding since fig- ures of speech usually deviate from their os- tensible meanings to express deeper semantic implications. Previous research lays empha- sis on the literary aspect of figures and seldom provide a comprehensive exploration from a view of computational linguistics. In this pa- per, we first propose the concept of figurative unit, which is the carrier of a figure. Then we select 12 types of figures commonly used in Chinese, and build a Chinese corpus for Contextualized Figure Recognition (ConFig- uRe). Different from previous token-level or sentence-level counterparts, ConFiguRe aims at extracting a figurative unit from discourse- level context, and classifying the figurative unit into the right figure type. On ConFiguRe, three tasks, i.e., figure extraction, figure type classifi- cation and figure recognition, are designed and the state-of-the-art techniques are utilized to implement the benchmarks. We conduct thor- ough experiments and show that all three tasks are challenging for existing models, thus re- quiring further research. Our dataset and code are publicly available at https://github. com/pku-tangent/ConFiguRe. 1 Introduction Figures of speech, also known as rhetoric figures or figurative languages, are a ubiquitous part of spoken and written discourse. These rhetorical techniques, such as metaphor, irony and paral- lelism, greatly enrich the expression of human lan- guages (Roberts and Kreuz, 1994). They intention- ally deviate from the literal meaning of language to provide deeper semantic expressiveness, therefore ∗Corresponding author †Author emails:{dwzhu, zhanqiusi, zhouzhejian, yfsong, zhangjiebin, lisujian}@pku.edu.cn Discourse Fragment: {…} [1] 这样的想了一想，[2] 我的断绝了联络的知觉，[3] 又重新恢复了转来， {[4] 一股同蒸气似的酸泪，} metaphor/simile（比喻） [5] 直涌了出来。[6] 我踉跄往后退了几步，[7] 倒在 外床她叠好在那里的那条被上。[8] 两手紧紧抱着 了这一条被，{[9] 我哭着，[10] 哭着，[11] 哭着， [12] 哭了一个尽情。} Parallelism（排比） {…} Translation: {…} Thinking of this, my disconnected senses started to restore again, [and a stream of sour tears gushed out like steam. ]metaphor/simile I stumbled back a few steps and fell out of the bed on the quilt she folded there. Holding this quilt tightly with both hands, [I cried, cried, cried and cried my heart out.]Parallelism {…} Figure 1: An annotated case of ConFiguRe. We present English translation in the second block for reading con- venience. Clause groups in italic are figurative units, with the specific figure type denoted in subscription. posing a big challenge to natural language under- standing. Linguists have a long history of studying rhetoric figures, extensively analyzing their use in literature, culture and psychology (Zhang, 1963; Wilks, 1975; Drew and Holt, 1998; Group, 2007; Shapin, 2012). These works mainly focus on collecting qualita- tive evidence. Figures of speech have also drawn attention from the NLP community. Many down- stream applications could be improved, should fig- ures be precisely identified and carefully dealt …

Content truncated to 500 words