Blow the Dog Whistle: A Chinese Dataset for Cant Understanding with Common Sense and World Knowledge

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 2139–2145 June 6–11, 2021. ©2021 Association for Computational Linguistics 2139 Blow the Dog Whistle: A Chinese Dataset for Cant Understanding with Common Sense and World Knowledge Canwen Xu1∗, Wangchunshu Zhou2∗, Tao Ge3, Ke Xu2, Julian McAuley1, Furu Wei3 1 University of California, San Diego 2 Beihang University 3 Microsoft Research Asia 1{cxu,jmcauley}@ucsd.edu 3{tage,fuwei}@microsoft.com 2zhouwangchunshu@buaa.edu.cn,kexu@nlsde.buaa.edu.cn Abstract Cant is important for understanding advertis- ing, comedies and dog-whistle politics. How- ever, computational research on cant is hin- dered by a lack of available datasets. In this paper, we propose a large and diverse Chinese dataset for creating and understanding cant from a computational linguistics perspective. We formulate a task for cant understanding and provide both quantitative and qualitative anal- ysis for tested word embedding similarity and pretrained language models. Experiments sug- gest that such a task requires deep language un- derstanding, common sense, and world knowl- edge and thus can be a good testbed for pre- trained language models and help models per- form better on other tasks.1 1 Introduction A cant2 (also known as doublespeak, cryptolect, ar- got, anti-language or secret language) is the jargon or language of a group, often employed to exclude or mislead people outside the group (McArthur et al., 2018). Cant is crucial for understanding ad- vertising (Dieterich, 1974) and both ancient and modern comedy (Sommerstein, 1999; Prasetyo, 2019). Also, it is the cornerstone for infamous dog- whistle politics (López, 2015; Albertson, 2015). Here, we summarize the key elements for cant: (1) Both a cant and its reference (i.e., hidden word) should be in the form of common natural text (not another symbol system, e.g., Morse code). (2) There is some shared information between the cant users (i.e., the insiders) that is not provided to the people outside the group. (3) A cant should be deceptive and remain undetected to avoid being Asia.∗Equal Contribution. Work done at Microsoft Research 1The code is available at https://github.com/ JetRunner/dogwhistle. The data and leaderboard are available at https://competitions.codalab.org/ competitions/30451. 2https://en.wikipedia.org/wiki/Cant_ (language) decrypted by people outside the group (i.e., the out- siders). These elements make the creation and un- derstanding of cant subtle and hard to observe (Tay- lor, 1974). To the best of our knowledge, currently there are very few resources available for the re- search of cant. In this paper, we create a dataset for studying cant, DogWhistle, centered around the afore- mentioned key elements (examples shown in Fig- ure 1). We collect the data with a well-designed online game under a player-versus-player setting (see Section 3.1). The dataset includes abundant and diverse cant for a wide spectrum of hidden words. We find that cant understanding requires a deep understanding of language, common sense and world knowledge, making it a good testbed for next-generation pretrained language models. Our dataset also serves as a timely and complex lan- guage resource that can help models perform better on other tasks through Intermediate Task …

Content truncated to 500 words