A Common Semantic Scoring Method for Chinese Subjective Questions

Abstract

With the rapid development of computer technologies and artificial intelligence, intelligent tutoring system is increasingly applied to our daily lives. This paper proposes a common semantic scoring method for Chinese subjective questions based on dependencies, modifiers and HowNet. First, we use dependencies to construct question classification predicate formulas for determining the question type and getting the characteristic words in the question. Then, we use dependency chains to extract multiple sets of score points in the answer according to the question type, and to optimize the answer's score points according to the feature words in the question sentence. Finally, we use the common semantic dictionary HowNet to calculate the similarities between the score points that have the same dependencies respectively in the student answers and the standard answer, and to combine the modifiers in answer sentences for calculating the final score of the subjective question. Experimental results show that our proposed method has the advantages of rapidity, accuracy and efficiency, and surpasses many excellent subjective question scoring methods.