COMPILING: A Benchmark Dataset for Chinese Complexity Controllable Definition Generation

CCL 2022 Proceedings of the 21st China National Conference on Computational Linguistics, pages 921-931, Nanchang, China, October 14 - 16, 2022. (c) Technical Committee on Computational Linguistics, Chinese Information Processing Society of China 921 COMPILING: A Benchmark Dataset for Chinese Complexity Controllable Definition Generation Jiaxin Yuan12*, Cunliang Kong123*, Chenhui Xie123, Liner Yang123† , Erhong Yang13 1National Language Resources Monitoring and Research Center Print Media Language Branch, Beijing Language and Culture University 2School of Information Science, Beijing Language and Culture University 3Beijing Advanced Innovation Center for Language Resources, Beijing Language and Culture University jiaxinyuan625@gmail.com Abstract The definition generation task aims to generate a word’s definition within a specific context au- tomatically. However, owing to the lack of datasets for different complexities, the definitions produced by models tend to keep the same complexity level. This paper proposes a novel task of generating definitions for a word with controllable complexity levels. Correspondingly, we in- troduce COMPILING, a dataset given detailed information about Chinese definitions, and each definition is labeled with its complexity levels. The COMPILING dataset includes 74,303 words and 106,882 definitions. To the best of our knowledge, it is the largest dataset of the Chinese definition generation task. We select various representative generation methods as baselines for this task and conduct evaluations, which illustrates that our dataset plays an outstanding role in assisting models in generating different complexity-level definitions. We believe that the COM- PILING dataset will benefit further research in complexity controllable definition generation. 1 Introduction Definition Generation (DG) is the task of describing the meaning that a word takes in a specific con- text. This task can help language learners by providing explanations for unfamiliar words. Recent re- searches (Ishiwatari et al., 2019; Zheng et al., 2021) attempted to apply the task to the field of Intelligent Computer-Assisted Language Learning (ICALL), and have made a significant progress. Previous studies on DG mainly concentrate on generating different definitions for polysemous words (Gadetsky et al., 2018; Mickus et al., 2019; Reid et al., 2020), or generating definitions with appropriate specificity (Huang et al., 2021a). In these studies, researchers have faced various issues, such as the high complexity problem. High complexity definitions contain words that are more difficult than the defined word, and hence are labored for language learners to read and understand. Nevertheless, there have been few focuses on complexity controllable generation of definitions. A possible reason is that the complexities of definitions are not provided in currently existed datasets, which leads to the difficulty of automatic training and evaluation. Actually, the problems mentioned above are especially prominent in the language environment of Chi- nese. Definitions with suitable complexity are in urgent practical needs for Chinese as Foreign Language (CFL) learners. According to the Ministry of Education of China, by the end of 2020, more than 20 million foreign students are learning Chinese. But as Zhang (2011) pointed out, since the difficulty of definitions is not considered, most existing dictionaries cannot meet CFL learner’s requirements. Be- sides, the existing Chinese learner dictionaries …

Content truncated to 500 words