TantyWidiyastuti/NLP-Hate-Speech-Detection

NLP Text Classification Benchmark

Natural Language Processing Project — Institut Teknologi Sumatera (ITERA)

Project Description

This project aims to compare the performance of Machine Learning (ML) and Deep Learning (DL) approaches for Natural Language Processing (NLP) tasks.

Machine Learning models will be implemented using PyCaret AutoML, while the Deep Learning model will be implemented using PyTorch.

Both approaches will be evaluated and compared on the same dataset to determine which method performs better for the selected text classification task.

Team Members

Dataset

Dataset used in this project will be sourced from public NLP datasets such as Kaggle, Hugging Face Datasets, or other open repositories.

Dataset Link: (https://github.com/okkyibrohim/id-multi-label-hate-speech-and-abusive-language-detection?tab=readme-ov-file)

Project Objectives

The objectives of this project are:

• —Perform Exploratory Data Analysis (EDA) on the selected dataset
• —Implement Machine Learning models using PyCaret AutoML
• —Implement Deep Learning models using PyTorch
• —Compare the performance between ML and DL models
• —Deploy interactive demos using Hugging Face Spaces
• —Publish a scientific report in ArXiv format

Repository Structure

pba2026-nama-kelompok 12
│
├── data
│   ├── raw
│   └── processed
│
├── notebooks
│   ├── 01_eda.ipynb
│   ├── 02_preprocessing.ipynb
│   ├── 03_pycaret_model.ipynb
│   └── 04_deep_learning.ipynb
│
├── src
│   ├── preprocessing.py
│   ├── train_ml.py
│   ├── train_dl.py
│   └── utils.py
│
├── models
│
├── app
│   ├── ml_demo
│   └── dl_demo
│
├── paper
│
└── README.md

Machine Learning Approach

Machine Learning models will be developed using PyCaret AutoML. Several algorithms will be compared automatically, and the best-performing model will be selected based on evaluation metrics.

Examples of algorithms evaluated:

• —Logistic Regression
• —Random Forest
• —Support Vector Machine
• —Gradient Boosting

Deep Learning Approach

The Deep Learning model will be implemented using PyTorch.

Possible architectures include:

• —LSTM
• —GRU
• —CNN for text classification
• —Lightweight Transformer models

The model will be trained and evaluated using standard NLP evaluation metrics.

Deployment

Two interactive demos will be deployed using Hugging Face Spaces:

• —Machine Learning Model (PyCaret) (link will be added later)

• —Deep Learning Model (PyTorch) (link will be added later)

Scientific Paper

The final project report will be written in LaTeX using ArXiv format and will include:

• —Dataset description
• —Methodology
• —Experiment setup
• —Benchmark results
• —Comparative analysis

ArXiv Link: (to be added)

Course Information

Course: Pemrosesan Bahasa Alami Program: Sains Data — Institut Teknologi Sumatera Semester: Genap 2025/2026

Instructor: Martin C.T. Manullang