Language-learning applications define exercises that are pedagogical tools to introduce new language concepts. The development of this type of applications is complex due to the diversity of language-learning methodologies, the variety of execution environments and the number of different technologies that can be used. This article proposes a complete Model-Driven Architecture (MDA) approach, from the definition of the Computational Independent Model (CIM layer) to the Implementation Specific Model (ISM layer), and the process of the necessary transformations for the automatic generation of the source code (in HTML and JavaScript) of language-learning applications. To carry out the model-to-model and model-to-text transformations, the ATLAS Transformation Language (ATL) and Acceleo transformation languages have been used respectively. The proposal has been validated through the modeling and the complete automatic generation of source code of two Learning Activity Mechanisms (LAM), which are used within methodologies such as Duolingo and Busuu: LAM Image-Audio-Text and LAM Audio-Text Options.

中文翻译：

---

语言学习应用程序的自动代码生成

语言学习应用程序定义了练习，这些练习是介绍新语言概念的教学工具。由于语言学习方法的多样性、执行环境的多样性以及可以使用的不同技术的数量，此类应用程序的开发非常复杂。本文提出了一个完整的模型驱动架构（MDA）方法，从计算独立模型（CIM 层）的定义到实现特定模型（ISM 层），以及源代码自动生成的必要转换过程语言学习应用程序的代码（HTML 和 JavaScript）。为了执行模型到模型和模型到文本的转换，分别使用了 ATLAS 转换语言 (ATL) 和 Acceleo 转换语言。