Thisisruiii/ag-news-topic-classification-processed

AG News Topic Classification Processed Dataset

This dataset is a processed subset of the AG News Classification Dataset from Kaggle.

Task

The task is news topic classification. Each example contains a news title and description combined into a single text field. The goal is to classify each news text into one of four categories:

• —World
• —Sports
• —Business
• —Sci/Tech

Data Source

The original data comes from the AG News Classification Dataset on Kaggle. For this assignment, I created a processed version for an embedding-based text classification project.

Processing

The original dataset contained the following columns:

• —Class Index
• —Title
• —Description

For this processed version:

1. 1.Combined Title and Description into one text column.
2. 2.Mapped numeric class labels to readable topic names.
3. 3.Created a balanced subset for training and testing.

Files

• —train_small.csv: 4000 examples
• —test_small.csv: 1000 examples

Columns

• —text: the combined news title and description
• —label: numeric class label
• —label_name: readable class name

Label Mapping

Intention

This dataset is used to train and evaluate a news topic classifier based on sentence embeddings. It was created for an Information Retrieval assignment involving embedding-based classification and a Hugging Face demo.