Dynamic Conversational Chatbot for Assessing Primary Students

Abstract

Teaching necessitates a method of determining if learners are gaining the desired knowledge and skills. We believe that chatbot technology would be an excellent solution to this problem. Using our approach, the chatbot will assess the student by using questions and answers recorded in a question bank. Four approaches were taken to assess the students’ answers against the model answer, using models such as Word2Vec, all-mpnet-base-v2, and Sense2Vec models with Word Mover’s Distance algorithm or Cosine Similarity. After evaluating these approaches, the best performing approach was when we used the Sense2Vec model with Cosine Similarity which gave the most accurate similarity score range for correct and incorrect answers.