How should we handle parent tags and tag trees?

Question

I've spent quite a bit of time thinking about the tagging structure we're using on this site, and one thing that particularly concerns me right now is parent tags or tag trees. Any tag which is a complete superset of another tag can be considered a parent tag, and when we get multiple levels of parent tags we end up with a tag tree.

What are parent tags and tag trees?

Here's a concrete example: What are the best sources of vitamin B12 for vegetarians?

This question is about vitamin B12, so of course b12 is used. Of course, applying a tag for any specific vitamin or mineral invites use of vitamins-and-minerals. But vitamins are just one facet of nutrition so we'll need to include that too. And the most common reason for thinking about nutrition is that we are concerned about our health.

The most detailed tag implies use of a more broadly scoped tag, and so on up the tree.

b12 → vitamins-and-minerals → nutrition → health

I contributed to the development of this problem by suggesting that we should have tags for specific vitamins and minerals. This confirmed a whole new layer that was already covered by the three other tags shown.

What's wrong with tag trees?

The biggest problem with tag trees is that when they get big, they start crowding out other useful tags. And if a tag tree ever grows to be 6 levels high (larger than the question tag limit), then anybody using the tag tree would need to decide which tag to omit: the least detailed, the most detailed, or break the chain in the middle.

Is there anything good about tag trees?

The Related Tags view looks great when a tag tree is in place and questions are completely tagged. Stack Exchange is also pretty good about showing related questions when tag trees are fully developed.

Do we have other tag trees?

We only have one big tag tree and it's underneath health.

• amino-acids → protein → nutrition → health
• cholesterol → fats → nutrition → health

We also have a couple smaller ones that have formed.

• dairy-substitutes → substitutes
• is-it-vegan → veganism
• honey → bees
• public-health → health
• crops → agriculture

This question evolved out of a brief discussion I had with Zanna in chat.

Answer 1

I found a bunch of semi-related questions on SO Meta but nothing that really answers this question. So instead I went looking to see what practices are currently in place on the trilogy sites. I figure if any tag hierarchies are in place, they will probably be visible on the first few pages.

Precedent

Stack Overflow

• pandas → python (87%)
• spring → java (59%)
• winapi → windows (59%)
• jquery → javascript (55%)

ServerFault

• puppetmaster → puppet (92%)
• bind → domain-name-system (64%)
• ubuntu → linux (22%)

SuperUser

• ext4 → linux (51%)
• nvidia-graphics-card → graphics-card (32%)
• windows-7 → windows (13%)

Perspective

Tag trees seem to be very rare on the trilogy sites, and definitely don't dominate the most popular tags like our health and nutrition tags do. The vast majority of cases appear to fit into a two-level relationship. There seems to be a strong preference towards using the most specific tags and ignoring the least specific ones, although I suspect that the size of each site is critical for determining what the most specific tags look like.

Conclusion

We should stop using overly broad tags and focus on keeping the hierarchy shallow.

It is absolutely not necessary to add parent tags when using more detailed tags. For example, not all questions tagged with b12 need to be tagged with vitamins-and-minerals also. If a parent tag is desired, then it's useful to have that tag mentioned on a few questions that are more specific so that related tags can be more easily discovered. By avoiding strict use of parent tags, tag trees become a lot less harmful.