Automatic assessment of short answers based on computational and data mining approaches

Automatic short answer grading is known as the task of automatic assessment of answers based on natural language using computation methods and machine learning algorithms. The proliferation of large-scale intelligent education systems and the importance of assessment as a key factor in the education process have increased the need for highly flexible automated systems for scoring exams.

While in the process of automatic short answer grading, student's answer is compared to an ideal response and scoring is done based on their similarity, semantic relatedness and similarity measures can also be employed for this aim. To this end, several semantic relatedness and similarity measures are firstly compared in application of short answer grading. In the following, a method for improving the performance of short answer grading systems based on semantic relatedness and similarity measures which leverages students' answers with the highest score as feedback is proposed.

In order to evaluate the performance of semantic and similarity relatedness methods in application of automatic short answer grading and the prposed model, various experiments were concucted on Mohler and Mihalcea dataset that contains 7 questions and 630 answers.

 Based on the empirical experiments not only semantic relatedness and similarity measures have great efficiency in automatic short answer grading but also using students' answers as feedback can considerably improve the accuracy and performance of semantic relatedness and similarity measures for this task.