Vietnamese-Text-Classification

A project for text classification in Vietnamese using the VNews8td dataset. The dataset and this repository are designed for document classification tasks.

Table of Contents

• Dataset: VNews8td
  • Dataset Details
  • Dataset Link
• Project Overview
• Installation
• Usage
• Contributing

---

Dataset: VNews8td

The VNews8td dataset is a Vietnamese text classification dataset collected from VnExpress from 01/06/2023 to 01/06/2024. It contains articles categorized into eight classes, with each article including a title and a description.

Dataset Details:

• Classes:

  • doisong (Đời sống)
  • giaoduc (Giáo dục)
  • khoahoc (Khoa học)
  • kinhte (Kinh tế)
  • suckhoe (Sức khỏe)
  • thegioi (Thế giới)
  • thethao (Thể thao)
  • thoisu (Thời sự)

• Splits:

  • Training Set: 70%
  • Validation Set: 10%
  • Test Set: 20%

Dataset Link:

Download the dataset

---

Project Overview

This project implements a text classification pipeline for Vietnamese documents using the VNews8td dataset. The primary steps include:

1. Preprocessing Vietnamese text using VnCoreNLP.
2. Extracting features with TF-IDF.
3. Training models such as Logistic Regression for classification.
4. Evaluating model performance with metrics like accuracy and classification reports.

---

Installation

1. Clone the repository:

   git clone https://github.com/thanghd1112/Vietnamese-Text-Classification.git
   cd Vietnamese-Text-Classification

2. Install required packages:

   pip install -r requirements.txt

3. Download and set up VnCoreNLP:

   wget https://github.com/vncorenlp/VnCoreNLP/archive/refs/heads/master.zip
   unzip master.zip
   rm -f master.zip
   mv VnCoreNLP-master VnCoreNLP

---

Usage

1. Preprocess and prepare the dataset:

   • Load the data from the provided link.
   • Split the data into training, validation, and test sets.

2. Train the model:

   • Use the Jupyter Notebook provided (VN_TextClassification.ipynb) to preprocess, train, and evaluate your model.

3. Evaluate:

   • Check accuracy and other metrics on the test set.

---

Contributing

Contributions are welcome! Feel free to submit a pull request or raise an issue if you encounter problems.

---