Was anything found that was not already well-known?

The article states:

"Wrap Up

... we’ve scraped hundreds of thousands of articles, created text classifiers... and gotten insights from the results. The only way to perform an analysis like this is using machine learning and natural language processing, since there’s no way we can get a person to read through and interpret 270,000 articles."

I would have to say

- there's always sampling, and

- ask or read the reports of an expert, since they live in their respective news spaces.

why asking an expert when you can ask the data?

Because you need an expert to make sense of the data? Which is why I am trying to transition to data science from sysadmin at the moment. So I don't need no stinkin middle men.

Articles can't be treated as data yet except with respect to very simple questions.

Why is machine learning/NLP needed in this analysis?

The categories/keywords are well known, and I believe that simple statistics would suffice to produce the type of results shown.

Of course, the text would have to be scanned and the dataset cleansed/filtered first. But those techniques predate ML/NLP.

50+% of the articles they found are the self-published like noise around fundraising and launch. Throws those obvious​ fluff out

Why publish this and not public the code they used to interface with MonkeyLearns API? Missed opportunity.

They published the code, you can check it out here: https://github.com/monkeylearn/startup-news-analysis/blob/ma...

This is a nice summary for those who haven't spent the last ten ten years in the startup sene, thanks!

Couldn't it actually mean the opposite? If fundraising is not newsworthy, it could also be because there are so many of those out there that fewer make the cut.

For start-ups without substance, yes. Fortunately.