Assessment of Indonesian Short Essay using Transfer Learning Siamese Dependency Tree-LSTM

The automatic short essay assessment is considering many aspects, not only the student's answer, but also the teacher marking rubrics for getting precise answers. Sometimes the marking rubrics are not used, instead, the teacher only used a single precise answer so that condition raises scoring ambiguity. This paper presents some precise answers for a fair assessment and discovers the important representation of the answer using grammatical dependency. Grammatical used to overcome uncertainty relation word in answer. The precise answers as the correct answer tolerance and student answers are used as the new input pair. Due to the Indonesian short essay is a lack, transfer learning is carried out. In this paper, we present a transfer learning using Siamese dependency tree-LSTM for obtaining the assessment score. The evaluation of the proposed method using Quadratic Weighted Kappa and accuracy that is 85.74% and 70.80%, respectively.