Recommendation System using XLNet and Graph based Reinforcement Learning for Japanese Language Teaching with Digital Resources

In today’s world, digital resources play a crucial role in the sector of language education, offering interactive sessions. In Japanese language teaching, digital resources such as multimedia content, apps, and virtual classrooms are employed rapidly to overcome the challenges such as complex Japanese writing systems, such as kanji, and handling cultural nuances within digital platforms and diverse learning requirements to improve language acquisition. This paper suggests a novel approach technique using various algorithms at each stage. The data is collected from Ilmjapanese-dataset. The collected data is fed as input to the XLNet transformer for processing of data since it has the ability to effectively analyse the text data and then the processed data is converted to vector using word2 vec technique. The overall data is transformed to graphical format and to enhance the overall system Graph Attention Networks v2 (GATv2) is incorporated to overcome the limitation of static attention models and in the end graph based reinforcement learning (GRL) is employed for developing a recommendation model for Japanese language teaching. The XLNet-GRL resulted with higher results with accuracy of 97.58 %, Specificity of 96.85 %, Sensitivity of 98.69 %, and F1 Score of 96.20%.