AG News Corpus with Corrected Text Labels

Recognai's corrected version of the AG News classification dataset, originally containing thousands of news articles. The dataset was updated on Hugging Face in December 2021. It focuses on text correction and classification tasks for machine learning.

Use Cases

• Train a classifier to predict the corrected news category label from the article text.
• Benchmark text correction algorithms by comparing original and corrected labels for a set of articles.
• Analyze label noise and correction patterns within the 'title' and 'description' fields of news articles.

Strengths

• Dataset is hosted on a major platform (Hugging Face) with a specific version history.
• Focuses on a known task (AG News classification) with an explicit quality improvement (corrected labels).

Limitations

• Unknown row count and column details limit assessment of scale and feature richness.
• Data is from 2021, potentially missing recent news trends and language.

Provenance

Source

Recognai (via Hugging Face).

Collection Method

Correction of labels from the original AG News dataset.

Time Range

null

Freshness

Last updated 2021-12-29.

Geography

Region tags suggest US focus.