AG News Article Classification Dataset

AG News is a text classification dataset for news articles. It is a popular benchmark for machine learning models, originally sourced from the AG's news corpus. The dataset is available on the OpenML platform.

Use Cases

• Classify news articles into categories like 'World', 'Sports', 'Business', and 'Sci/Tech' using the article text field.
• Train and evaluate text classifiers, such as naive Bayes or neural networks, on the provided article titles and descriptions.
• Analyze word frequency and topic modeling on the corpus of news article text.
• Benchmark model performance on a standard text classification task against published results.

Strengths

• Standard benchmark dataset for text classification tasks.
• Well-defined classification categories for evaluation.

Limitations

• Specific row count, column details, and temporal coverage are unknown.
• Potential class imbalance or geographic bias in the original news corpus is not specified.

Provenance

Source

OpenML platform, originally from AG's news corpus.

Collection Method

Articles collected and categorized from news sources.