Fake and Real News Articles with Linguistic and Credibility Features

A collection of news articles labeled as fake or real, annotated with linguistic and credibility features. The dataset's size, author, and last update date are unknown. It was sourced from the Kaggle platform.

Use Cases

• Train a binary classifier to predict the 'fake' or 'real' label using linguistic features like word count and sentiment scores.
• Analyze the correlation between specific credibility features, such as source ratings or claim verifications, and the article's truth label.
• Build a model to identify fake news by extracting stylistic patterns from the article text content.
• Conduct feature importance analysis to determine which linguistic attributes (e.g., punctuation frequency, readability scores) are most indicative of fabricated content.

Strengths

• Includes explicit linguistic and credibility feature annotations for model training.
• Provides a clear binary classification task (fake vs. real) for supervised learning.

Limitations

• The total number of articles (rows) is unknown, making it difficult to assess statistical power.
• Potential geographic and temporal bias exists as the source and time range of the collected news are unspecified.
• Class imbalance between fake and real news samples could affect model performance if not addressed.

Provenance

Source

Kaggle

Collection Method

null

Time Range

null

Freshness

null

Geography

null