View Proposal

---

Proposer

Ian Tan

Title

Automated Essay Grading using Fine-Grained Linguistic Features

Goal

Extracting fine-grain linguistic features, conduct dimensionality reduction, and applying classification algorithms to grade essays.

Description

The intricate relationship between fine-grained linguistic features and writing quality is to be reproduced. This is based on a paper titled "Incorporating Fine-Grained Linguistic Features and Explainable AI into Multi-Dimensional Automated Writing Assessment" by Tang et al. (2024). The paper harnessed computational analytic tools and Principal Component Analysis (PCA) to distill and refine linguistic indicators for model building and construction. This project is to reproduce the work done by Tang et al. and is to be scoped accordingly, likely to be with a subset of the feature extraction tool, and without the explainable AI component. A subset of the features is hypothesised as the number of features used for machine learning has diminishing returns as features don’t all contribute equally. A small subset of features explains most of the variance in the essay score and as more features are added, their marginal contribution (% influence) drops significantly. Also, having too many features with low influence are essentially noise, and including too many weak features increases complexity (computational cost penalty), and risks overfitting.

Resources

Some open sourced tools to be determined. Python programming.

Background

Main reference: * Tang, X., Chen, H., Lin, D. and Li, K., 2024. Incorporating fine-grained linguistic features and explainable ai into multi-dimensional automated writing assessment. Applied Sciences, 14(10), p.4182. Using LLM to grade: * Liew, P.Y. and Tan, I.K., 2024, December. On automated essay grading using large language models. In Proceedings of the 2024 8th international conference on computer science and artificial intelligence (pp. 204-211). Dataset Distribution: * Tan, J.S., Tan, I.K., Soon, L.K. and Ong, H.F., 2022. Improved automated essay scoring using gaussian multi-class SMOTE for dataset sampling. In Proceedings of the 15th International Conference on Educational Data Mining (p. 647).

Url

External Link

Difficulty Level

Moderate

Ethical Approval

None

Number Of Students

1

Supervisor

Ian Tan

Keywords

essay grading, machine learning, dimensionality reduction, feature extraction

Degrees

Bachelor of Science in Computing Science