StyleKQC: A Style-Variant Paraphrase Corpus for Korean Questions and Commands

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 7122–7128 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 7122 StyleKQC: A Style-Variant Paraphrase Corpus for Korean Questions and Commands Won Ik Cho1, Sangwhan Moon2, Jong In Kim3, Seok Min Kim1, Nam Soo Kim1 Department of Electrical and Computer Engineering and INMC, Seoul National University1 Department of Computer Science, Tokyo Institute of Technology2 Interdisciplinary Program in Cognitive Science, Seoul National University3 wicho@hi.snu.ac.kr,sangwhan@iki.fi,prows12@gmail.com smkim@hi.snu.ac.kr,nkim@snu.ac.kr Abstract Paraphrasing is often performed with less concern for controlled style conversion. Especially for questions and commands, style-variant paraphrasing can be crucial in tone and manner, which also matters with industrial applications such as dialog systems. In this paper, we attack this issue with a corpus construction scheme that simultaneously considers the core content and style of directives, namely intent and formality, for the Korean language. Utilizing manually generated natural language queries on six daily topics, we expand the corpus to formal and informal sentences by human rewriting and transferring. We verify the validity and industrial applicability of our approach by checking the adequate classification and inference performance that fit with conventional fine-tuning approaches, at the same time proposing a supervised formality transfer task. Keywords: Paraphrase, Style-variant, Korean, Spoken language, Directives 1. Introduction Paraphrasing, the act of using different sentences with the same meaning (Bhagat and Hovy, 2013), is strongly related to the text style conversion or transfer (Yamshchikov et al., 2020). While prior studies often modify sentiment or offensiveness (Logeswaran et al., 2018; dos Santos et al., 2018), in view of paraphrasing, it should be well checked whether the core content of the sentence is maintained during the conversion process. If the sentence meaning stays the same while changing politeness or formality (Rao and Tetreault, 2018), we can call it paraphrasing or rewriting. Such styles can be represented in diverse ways across genre, domain, and language (Jhamtani et al., 2017; Fu et al., 2018; Yang et al., 2019). Up to date, paraphrasing has been adopted as a use- ful strategy for text data augmentation. For instance, recent text augmentation schemes such as Dhole et al. (2021) exploit various automatic rewriting tools such as abbreviating, transliteration, lexical shift etc., while paraphrasing through text style transfer is one of them. The approach bases on unsupervised learning of En- glish text styles, following the scheme of Krishna et al. (2020). However, automatic style transfer may not always guarantee the naturalness of the sentence and the preservation of core contents. Also, it is not easy to attain direct text style transfer pairs from unsupervised and automatic approaches. This challenge is visible in the languages with a comparably lower amount of re- sources, where substantial resources are not guaranteed for each desired text style. In this light, we attempt to make up a solid scheme for the manual construction of text style transfer database, in a less studied language, Korean. We deal with the scheme of constructing a corpus of style-variant para- phrases …

Content truncated to 500 words