I wouldn't trust keyword analysis or word frequency to figure out why people are leaving. And unless AI has gotten a lot smarter than I'm aware of, I'd say that the only way to get a meaningful result is to actually read each blog post and then write a short 5-10 word summary. Sure, it's subjective but you'd get a lot more nuance like, "His girlfriend lives in New York", "Misses skiing and now has money to do just that", "Closer to elderly parents who need care", "New startup in Texas that's perfect fit", "Can't renew U.S. visa", "Moving to attend grad school".

The author has pared the list down to 137 blog posts, so even a fast reader would take a day to read all of them. But probably he already spent at least a day collecting that data, doing the analysis, creating the graphics, etc., so it's not crazy to suggest doing it manually. Alternatively, the list could be further cut down to 20-40 postings, which could be read in a few hours. That would result in a way more interesting (human) analysis.

EDIT: For each blog post, it would be interesting to note the sentiment of whether the person is being pushed out of San Francisco (can't afford the city, lost job, can't renew visa, etc.) or whether they are being pulled to a new location (perfect new job, girlfriend/boyfriend in different location, etc.).