It is widely understood that Artificial Intelligence (AI) could potentially be applied to offer more effective education in a large scale by providing students with a learning experience that is tailored towards their individual needs. To realize individualized education, it is required that the AI be capable of assessing the learner’s current knowledge level, evaluating the difficulty level of the learning materials, and finding the most appropriate learning tasks based on the learners’ individual differences. In other words, the construction of an Intelligent Tutoring System (ITS) requires valid models for the education domain, the learner, and the tutoring process, which all require an effective method to represent the knowledge being taught. There has been a plethora of research on ITSs and knowledge representation for Science, Technology, Engineering, and Mathematics (STEM) subjects. However, much less has been done in the language education domain, mainly due to reasons of research tradition, complexity of language knowledge representation, and the high theoretical and technical barriers. As a result, the current project aims at creating a technology for Second Language Knowledge Representation (L2KR) by making use of Natural Language Processing (NLP) and other AI techniques. This L2KR technology will not only be applicable to adaptive input selection, adaptive feedback generation, and adaptive pedagogical decision making in L2 education research, but it will also be able to feed back to improve the technology of controllable Natural Language Generation (NLG). The L2KR technology will make it possible to control the readability of the generated language so as to make it more easily understandable by target users, especially those with lower language proficiency.
We propose to develop the L2KR technology with a pedagogical orientation to make the decision making process in future intelligent language tutoring systems more transparent, explainable, and easily understandable, thus also increase the acceptability of the technology among language educators. The readability controllable NLG technology adds a new dimension of control to current NLG technology, potentially increasing the technology’s acceptability and usefulness.