Financial News Multisource

57.1 million rows of financial and general news content aggregated from 24 public datasets by Brianferrell787, last updated in November 2025. The corpus is normalized into a consistent schema and provided in Parquet format for high-performance querying and streaming.

Use Cases

• Language modeling using the 'text' column for domain-specific pre-training
• News summarization tasks using the 'text' content to generate headlines or abstracts
• Market sentiment analysis by joining the 'date' column with historical financial time-series data
• Topic classification using the 'text' and 'extra_fields' columns to categorize news by sector

Strengths

• 57.1 million rows of content
• Aggregates 24 distinct public news datasets into one layer
• Normalized schema across all sources
• Parquet format optimized for DuckDB and Polars

Limitations

• Minimal schema (date, text, extra_fields) may omit source-specific granular metadata
• Potential for duplicate articles across the 24 source datasets
• Geographic bias toward the US as indicated by metadata tags

Provenance

Source

24 public news datasets (unspecified names)

Collection Method

Aggregation and normalization of existing public datasets

Freshness

Last updated November 2025; update frequency is not specified.

Geography

United States