Automatic Short Answer Grading With SemSpace Sense Vectors and MaLSTM

Automatic assessment of exams is widely preferred by educators than multiple-choice because its efficiency in measuring student performance, lack subjectivity when evaluating response, and faster evaluation time the consuming manual evaluation. In this study, a new approach for Short Answer Grading (ASAG) proposed using MaLSTM sense vectors obtained SemSpace, synset based embedding method built leveraging WordNet. Synset representations Student's answers reference are given as input into parallel LSTM architecture, they transformed sentence hidden layer vectorial similarity these two representation computed with Manhattan Similarity output layer. The has been tested Mohler ASAG dataset successful results terms Pearson (r) correlation RMSE. Also, case study specific (CU-NLP) created from exam “Natural Language Processing” course Computer Engineering Department Cukurova University. And it achieved correlation. experiments show that system can be used efficiently effectively context-dependent tasks.