AUTOMATED SHORT-ANSWER GRADING USING SEMANTIC SIMILARITY AND SYNTACTIC ANALYSIS TO DETECT WORD ORDER IN ANSWER SENTENCE

• Main Author: Efendi, Toni
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71402
• Institution: Institut Teknologi Bandung

Essay tests are considered capable of measuring students' complex abilities, such as freedom in answering and formulating ideas. However, they have weaknesses; they require a longer assessment time and a high degree of subjectivity to produce different assessments. Therefore, automatic short answer grader (ASAG) is needed to help the assessment process faster and more objective. Most of the research related to ASAG focuses on increasing accuracy to approach the results of the manual assessment by humans. Several methods can improve ASAG accuracy, one of which is by increasing the number of corpus texts used as input for training the word embedding model. This model can anticipate various student answers by mapping words with close meanings. However, the word embedding model cannot provide precise judgments like humans when matching answers that require word order accuracy. Therefore, this study aimed to add syntactic analysis. The syntactic analysis utilizes the part-of-speech (POS) method and dependency parsing to detect word order in the answer sentences. The test results show an increased accuracy, which is better than previous research. The developed model achieved a correlation coefficient value of 0.6957 and a mean absolute error value of 0.7257. Besides, the addition of syntactic analysis to detect word order in answer sentences has been successfully implemented. It is proven to increase the accuracy of the model in assessing essay answers automatically.