How are "Suggested Tags" chosen?

Question

After you've typed in a question, you get suggested tags that appear that look like this:

It appears that you need a minimum amount of characters (much like the limit on 'Similar Questions' in the right-side pane):

you must have entered a title (any length) and a minimum body length of 220 characters.

But other than that, how are the tags chosen? Is it:

1. A match of the title/body with other questions (taking the tags from those questions)
2. A match of the title/body with tag excerpts (taking tags with overlap)
3. A match of the title/body with tag wikis (taking tags with overlap)
4. All of the Above
5. None of the Above
6. Other (please explain)

Some additional info from a previous feature request:

The suggested tags seem to only pop up on meta.so (they do not work on regular Stack Overflow, or on other SE sites).

While we're on the topic, why isn't this feature implemented network-wide?

Answer 1

I caught balpha in the Tavern and he will likely give a full answer later, but for those of you who are dying to have even the slightest inkling on how this works you can read the above link, or the summary below:

1. It uses similar-to-Bayesian reasoning on the text from the title and body of a post
2. It has been trained on older questions
3. A minimum of around 20k questions are required for accurate results
4. There is no substitute for material to learn from, so it can't be rolled out to all sites

I'm sure there's more to it than that, but this may help those of you looking for information and/or plotting the classification of Meta.