User Adaptive Language Learning Chatbots with a Curriculum

Along with the development of systems for natural language understanding and generation, dialog have been widely adopted learning practicing. Many current educational perform chitchat, where generated content vocabulary are not constrained. However, learners in a school setting, practice through is more effective if it aligns students’ curriculum focuses on textbook vocabulary. Therefore, we adapt lexically constrained decoding to system, which urges system include curriculum-aligned words phrases its utterances. We adopt generative BlenderBot3, as our backbone model evaluate curriculum-based middle students English their second language. The derived from textbooks, suggested by teachers. evaluation result demonstrates that infusion improves target increases interest practicing English.