Datasets:
Upcoming additions to Financial-News-Multisource (12 new datasets) — feedback welcome
Hi all — quick update on the Financial-News-Multisource dataset.
This corpus is actively maintained. I’m integrating 12 additional datasets that expand recency (through 2025), add finance-specific sources (with tickers and timestamps), broaden general-news context for market events, include social discussion signals, and deepen tech-company coverage. As always, the corpus will be deduplicated, schema-consistent (date, text, extra_fields), and aligned to our research-only usage terms.
Why these additions help
- Recency & continuity: Yahoo Finance (ongoing 2025) + NYT 2000–present help keep coverage fresh and continuous.
- Finance focus: Jeet’s U.S. financial news and FinSen add stock-linked news and sentiment metadata.
- General news context: All-the-News and Frontpage broaden macro/political coverage impacting markets.
- Social discussion: Reddit finance posts capture retail investor discourse for robustness.
- Tech domain: HackerNoon significantly increases tech-company news volume in 2022–2023.
Planned additions (links):
- https://huggingface.co/datasets/bwzheng2010/yahoo-finance-data
- https://huggingface.co/datasets/AndyReas/frontpage-news
- https://github.com/felixdrinkall/financial-news-dataset
- https://www.kaggle.com/datasets/eaglewhl/finsen-financial-sentiment-dataset
- https://huggingface.co/datasets/emilpartow/reddit_finance_posts_sp500
- https://www.kaggle.com/datasets/aryansingh0909/nyt-articles-21m-2000-present
- https://www.kaggle.com/datasets/jeet2016/us-financial-news-articles
- https://components.one/datasets/all-the-news-2-news-articles-dataset
- https://www.kaggle.com/datasets/julianschelb/newsdata
- https://www.kaggle.com/datasets/jonasbecker98/news-website-articles
- https://www.kaggle.com/datasets/parsonsandrew1/nytimes-article-lead-paragraphs-18512017
- https://huggingface.co/datasets/HackerNoon/tech-company-news-data-dump
Feedback welcome: If you have suggestions (e.g., preferred filters, additional sources, field/schema tweaks), please comment here or open an issue. The goal is to keep this dataset broadly useful for non-commercial, research-only work in financial NLP.
Thanks for using and supporting the project!