Yea, I think a more interesting analysis would have based on frequency of profession. That is, how far away (+% / -%) from a random matching is it?

Was thinking the same. Something like pointwise mutual information would do the trick.

Yeah, this vis needs to filter out teachers and nurses since they are top results for anybody, and/or professions need to be color coded by category so I can glance at a list of 15 items and see that ${medicalprofessional} surprisingly marries educators.

Rather than filter just those two, it could have an option to re-weight by availability.

I'm also curious what percentage of people married to people in their career field is due to meeting on the job. Are you more likely to marry someone in your profession? Or is "working together" just a very common way to meet your spouse?