SHORT ANSWER GRADING MODEL USING TEXT SIMILARITY AND MACHINE LEARNING
The assessment of short answer answers by humans often encounters several problems. There are human error and inconsistency in assessing answers due to bias, fatigue or the other and lack of effectiveness, efficiency of assessors when exams/tests/assessments are on a large scale. Assessment of sh...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/55033 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The assessment of short answer answers by humans often encounters several
problems. There are human error and inconsistency in assessing answers due to
bias, fatigue or the other and lack of effectiveness, efficiency of assessors when
exams/tests/assessments are on a large scale. Assessment of short answers using a
computer machine is an alternative solution in dealing with this problem.
This study uses a machine learning model with decision tree to conduct an
assessment. The feature used in the building model is the text similarity by cosine
similarity based on Bag Of Word (BoW) and Latent Semantic Indexing (LSI).
Cosine similarity with BoW measures similarity in terms of lexical (related to
words). Meanwhile, cosine similarity with LSI measures the similarity in terms of
semantics (meaning of the word). The text similarity measure calculates the level of
similarity between student’s answers and the answer key. From these two features,
the student’s answer scores will be predicted. The performance of the model is
evaluated with a measure of accuracy. The accuracy of the best model obtained in
this study is 83%. |
|---|