---
license: apache-2.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: label
    dtype: int64
  - name: title
    dtype: string
  - name: description
    dtype: string
  - name: text
    dtype: string
  - name: text_vi
    dtype: string
  splits:
  - name: train
    num_bytes: 94742232
    num_examples: 120000
  download_size: 55392570
  dataset_size: 94742232
task_categories:
- text-classification
- token-classification
language:
- vi
- en
pretty_name: Vietnamese AG News
size_categories:
- 100K<n<1M
---

# Vietnamese AG News (Machine Translated)

## Overview
This is a Vietnamese version of the **AG News** dataset, created for experimental purposes and personal learning. I used a machine translation model to convert the original English content into Vietnamese to explore how translation affects text classification tasks.

## Translation Method
The translation was performed using the **`facebook/nllb-200-distilled-1.3B`** model.

* **Source**: AG News dataset (English)
* **Target**: Vietnamese

## Dataset Structure
The dataset contains the following columns:

- `label`: The category ID (0: World, 1: Sports, 2: Business, 3: Sci/Tech).
- `title`: The original English title.
- `description`: The original English description.
- `text`: A combination of `title` and `description` in English.
- `text_vi`: The machine-translated version of the `text` column in Vietnamese.

## Purpose
This project is strictly for fun and educational exploration. Please note that since the translation is automated, some nuances or technical terms may not be perfectly accurate.