Evaluating the Impact of Sentence Tokenization on Indonesian Automated Essay Scoring Using Pretrained Sentence Embeddings
Bibliographic record
Abstract
Automated Essay Scoring (AES) systems are designed to expedite the assessment process, where human scoring is frequently slow and subject to inconsistencies and inaccuracies. This study, therefore, investigates the role of sentence tokenization in the performance of Indonesian Automated Essay Scoring, given that Natural Language Processing (NLP) techniques are requisite in AES to handle student responses that present identical semantic meanings but vary in length. A distinct approach was adopted in which full answers were not vectorized; instead, they were fragmented into sentences prior to vectorization. This method was deemed potentially more effective due to the high probability of discrepancies in sentence order between reference and student responses. Sentence embeddings, which encapsulate a sentence as a sole vector, were utilized. Pretrained SBERT-based sentence embeddings were employed to vectorize sentences from both reference answers and student responses, serving as semantic features for the Siamese Manhattan LSTM (MaLSTM) model. The MaLSTM model possesses the ability to process two inputs and evaluate their similarity using the Manhattan distance metric and use this similarity value as a predictive scoring output. This score was subsequently compared to human scores using the Root Mean Square Error (RMSE) and Pearson Correlation. Interestingly, sentence embeddings without tokenization slightly outperformed those with sentence splitting, as evidenced by a 0.61% improvement in RMSE and a 0.01 increase in Pearson Correlation. The results obtained indicate that sentence tokenization, as applied to the Indonesian Automated Essay Scoring dataset, does not have a notable impact on essay scoring performance. Therefore, it may be concluded that the application of sentence tokenization is not a necessary step in this dataset's text-processing phase of AES.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.015 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".