ProQuest is the digital provider for Wall Street Journal historical archives. Many public libraries provide free access; often all you need is a local library card and login.

WSJ has a link to its digital archive, which only goes back to May 1996. And unfortunately it doesn't have API so getting articles out of site for NLP is a pain. Let me know if you have code for that. I'm working on something similar

Good idea.

The market sentiment in 1921 was very bad for stocks. In fact, a week ago in 1921, Andrew Mellon (Treasury Secretary) suggested retail investors avoid stocks. This was covered in my post last week.

Sentiment is going to flip around 1923-1924 (once we clear the 1920 top) and only strengthen into 1929.

@roaring20s I have read through all of your current Substack articles and would love to catch up on your own background and data sources. I think pre-computer data extraction powers the best ML insights.

I am working at a YC Fintech company exploring economic opportunity over time and think there are some deep relationships here around human psychology and markets.

Interesting - What company?

If you can't say, no worries.