Development of SQL Similarity with Multiple Answer Keys for the Automated Assessment Process in the SQLearn Application

Automated assessment plays an important role in computer science education especially in basic database subjects. Using automated assessment decreased the workload of a lecturer and increase student participation and engagement while learning database. In this study we propose to use a multiple answer key and similarity check using tf-idf and cosine similarity in our SQLearn automated assessment system. The experiment involves 20 sophomore computer science students that learn to create queries using SQLearn system. From this experiment we obtain 453 queries created by students, we use SQLearn to automatically assess the student’s answers using both single key answer and multiple key answers to measure the accuracy of SQLearn. By adding the multiple keys answer the SQLearn shows an optimum accuracy of 92.20% in 60% to 70% similarity threshold, this accuracy was increased 7.8% compared to single key system. The results also shown that the accuracy increased following the increased of provided key answer in this experiment we recommend minimum three answer keys for one question. By utilizing multiple answer key similarity check in SQLearn we can improve the accuracy of the assessment while reducing lecturer workload.