Hybrid Clustering Algorithm for Language Proficiency Profiling in Online Tutoring Systems
Abstract
The online tutoring systems are vital in improving language learning through the delivery of the personal content and dynamic feedback. First, the language proficiency of learners needs to be profiled accurately to be able to adjust the instructional approach and enhance the learning process. The available methodologies would tend to use standardized tests or simplistic clustering and these cannot demonstrate the multidimensionality of learner behaviour and performance thus resulting in inaccurate or rough profiling. This paper attempts to overcome these challenges by suggesting a Spectral-Spatial Clustering (SSC) framework that combines spectral analysis to obtain similarity patterns globally with spatial clustering to obtain the local structure. The hybrid nature of this approach allows the system to divide the learners into a more accurate category, according to performance indicators, interaction patterns and learning progress. The suggested approach enables online tutoring systems to create dynamic and granular proficiency representations and enables the use of this information to deliver content and provide adaptive learning solutions. Experimental testing proves that SSC is accurate and brings about significant learner splitting in comparison to conventional technologies.