NewsWire: 2.7 Million Georeferenced U.S. News Articles (1878-1977)

NewsWire contains 2.7 million unique public domain U.S. news wire articles published between 1878 and 1977. Created by Dell Research Harvard, the collection features georeferenced locations, neural topic classification, and named entities disambiguated to Wikipedia. It provides a century-long record of American news history with structured metadata for computational social science.

Use Cases

• Historical text generation using the 1878-1977 news corpus
• Spatial analysis of historical events using georeferenced location metadata
• Entity-centric research using Wikipedia-linked individual identifiers
• Topic modeling and classification of news trends over a century

Strengths

• 2.7 million unique records
• 100-year temporal coverage (1878-1977)
• Expert-modeled entity disambiguation to Wikipedia
• Georeferenced location data for spatial analysis

Limitations

• Geographic bias toward the United States
• Topic tags are generated by neural models and may reflect algorithmic bias
• Temporal reporting bias inherent in historical news wire services

Provenance

Source

Dell Research Harvard

Collection Method

Aggregated from public domain news wires and processed with neural topic classification and entity disambiguation models.

Time Range

1878-1977

Freshness

Static historical collection; last updated August 2025.

Geography

United States