---
license: apache-2.0
language:
- en
tags:
- news
- nlp
- titles
- abstract
- nyt
- corpus
- headlines
- pre-training
pretty_name: New York Times News 100 Years
size_categories:
- 10M<n<100M
---
# New York Times 100 Years of News Headlines (1927-2026)

This dataset contains approximately 100 years of New York Times news headlines and abstracts, ranging from **1927 to January 2026**. It is designed for time-series analysis, NLP tasks, and historical research.

**Hugging Face Dataset Page:** [bguzzo2k/nyt_100y_news_headlines](https://huggingface.co/datasets/bguzzo2k/nyt_100y_news_headlines)

## Dataset Description

The dataset consists of metadata for articles published by The New York Times. It captures the "Main" headline and the "Abstract" (summary) provided by the NYT Archive.

- **Sources:** [New York Times Archive API](https://developer.nytimes.com/docs/archive-product/1/overview).
- **Temporal Coverage:** January 1, 1927, to January 31, 2026.
- **Content:** Headline (Title) and Abstract (Summary) of articles.

## Data Format

The data is organized into yearly Parquet files located in the `yearly_parquet/` directory (e.g., `nyt_1927.parquet`, `nyt_2026.parquet`). 

Each file is optimized for performance using:
- **Compression:** Snappy
- **Indexing:** Sorted by publication date.

### Schema

| Column | Type | Description |
| :--- | :--- | :--- |
| `date` | datetime64[ns, UTC] | The publication date and time of the article. |
| `headline` | string | The main headline of the article. |
| `abstract` | string | A brief summary or abstract of the article content. |

## Data Generation & Maintenance

The dataset is generated and updated using the Python scripts found in the `utils/` directory:

1.  **`utils/nyt_archive_downloader.py`**: A multi-threaded downloader that fetches raw JSON data from the NYT Archive API. It uses exponential backoff to respect API limits.
2.  **`utils/nyt_pqt_convert.py`**: Processes the raw JSON files, extracts relevant fields, cleans the data, and converts it into structured Parquet files grouped by year.

## Usage

To read the data using Python and Pandas:

```python
import pandas as pd

# Example: Read the year 2025
df = pd.read_parquet("yearly_parquet/nyt_2025.parquet")
print(df.head())
```

## License

This dataset is provided for educational and research purposes. Please refer to the [New York Times Developer Terms of Service](https://developer.nytimes.com/terms) regarding the use of their API data.