Auto-Grading Comprehension on Reference-Student Answer Pairs using the Siamese-based Transformer

Students often struggle with comprehension skills, which are critical for success in many areas of life. To sustain the growing demand for online education setups requires sophisticated evaluation tools, beyond meremultiple-choice format to effectively groom students. The research work proposes RC-based grading tools using the Siamese-based Transformer modelon reference-student answer pairs as the reference-based approach. The technique relies on creating a custom Representation QA model as Bi-Encoders in the Siamese-based transformer model. Furthermore, the research work presents the experimental study of the potential of using LLM such as ChatGPT to replace human annotation of reference answers by synthesizing through provided passages and questions. The research work is presented on the standard mocha benchmark dataset.