📝 Automated Essay Scoring (AES) – Kaggle Project
Automated Essay Scoring (AES) is an intelligent system designed to evaluate written essays using machine learning and deep learning techniques. This project focuses on developing a reliable scoring model using the Kaggle AES dataset, applying modern NLP preprocessing, feature engineering, and state-of-the-art neural architectures. The goal is to create a scoring pipeline that performs consistently, reduces human bias, and delivers accurate, fast, and scalable evaluations of student writing.
📚 Table of Contents
📖 Overview
📁 Repository Structure🎯 Objectives
🚀 Features
🛠️ Installation
📊 Usage
🧹 Data Preprocessing
🧠 Models Used
📈 Evaluation
📂 Dataset
📚 References
📝 License
📖 Overview
Automated Essay Scoring (AES) is transforming the way learning assessments are conducted by improving the speed, consistency, and fairness of scoring. Manual grading is slow and labor-intensive, especially in areas with limited educational resources. AES applies machine learning and NLP techniques to automate the scoring process.
This project provides a comparative study of multiple AES models trained on one of the largest publicly accessible essay datasets aligned with modern educational standards.
Model Performance Summary Model Cohen’s Kappa Linear Regression 0.6540 XGBoost 0.7100 LightGBM (LGBM) 0.7210 LSTM 0.7710 BERT 0.7806
Deep learning models — especially LSTM and BERT — significantly outperform traditional machine learning methods in prediction accuracy and reproducibility.
The goal of this work is to propose a publicly available AES system that improves teacher workflow efficiency and provides students with fast, objective feedback.
📁 Repository Structure
├── Dataset │ ├── train.csv │ ├── test.csv │ ├── Notebook │ ├── AES_model.ipynb │ ├── AES_inference.py │ ├── References │ ├── research_paper.pdf │ ├── .gitignore ├── LICENSE ├── README.md ├── requirements.txt
🎯 Objectives
1.Automate essay scoring using ML and NLP
2.Compare traditional and deep learning approaches
3.Extract linguistic & semantic features
4.Build train-ready and inference-ready scripts
5.Improve feedback speed for students
🚀 Features
✔ Full AES pipeline (preprocessing → training → prediction) ✔ Multiple model comparisons ✔ Deep learning support (LSTM, BERT) ✔ Clean notebook + Python script ✔ Requirements.txt for easy setup ✔ MIT licensed
🛠️ Installation
1️⃣ Clone the repository
git clone https://github.com/YourUsername/Automated-Essay-Scoring.git cd Automated-Essay-Scoring
2️⃣ Install dependencies
pip install -r requirements.txt
📊 Usage
Run Jupyter Notebook jupyter notebook
Open file: Notebook/AES_model.ipynb
Run Python Inference Script python Notebook/AES_inference.py
🧹 Data Preprocessing
The pipeline includes:
1.Lowercasing
2.Punctuation removal
3.Stopword removal
4.Lemmatization
5.Tokenization
6.Word & sentence statistics
7.Readability metrics
8.TF-IDF feature extraction
🧠 Models Used
Model Type Description Linear Regression Baseline scoring model RandomForest / XGBoost / LGBM Strong for tabular + text stats LSTM Sequential deep learning model BERT Transformer-based, best semantic understanding
📈 Evaluation
Primary Metric:
Used in Kaggle AES competitions to measure agreement between model predictions and human graders.
📂 Dataset
Dataset contains:
Essay text
Essay set ID
Human-graded score
You must download the dataset from Kaggle: 👉 Kaggle → Automated Essay Scoring Dataset(https://www.kaggle.com/competitions/learning-agency-lab-automated-essay-scoring-2/data)
Size: 24,000 argumentative essays
Score Range: 1 to 6
📚 References
Kaggle AES Competition
NLP Educational Research
Included Research PDF in /References/
📝 License
This project is licensed under the MIT License. See the LICENSE file.