In this long post, I will describe a few issues associated with tags that I believe should be solved. Feel free to add your opinions regarding the issues or specific tags (or feel free to edit this question to add more issues or tags)

Tags that should NOT exist

I've observed that, on this website, there are a lot of tags whose meaning is not widely accepted yet or it is not understandable. For example, . I've never heard of this expression before (and I am involved in the field of AI). Even though an expression occurs in a web article, it doesn't mean that it deserves a tag on this website.

I believe that only topics that have a "considerable amount" of associated research literature should have an associated tag, for example, questions related to ant colony optimisation () or related to or RL algorithms.

There are other tags that I believe should not exist on this website:

In general, as a rule of thumb, if a topic or expression does not have an associated Wikipedia article, it likely means that a corresponding tag should not exist on this website. I have not given a motivation for each of the tags, because otherwise I would not do anything more today. It is possible that in some cases I am not seeing the reason of the existence of the corresponding tags. However, I believe that most of them should not exist.


There are tags that are duplicates

  • and (I would leave just first one, given that the first tag is more "standard", even though we can have synonymous tags)

  • and (I would say that they can co-exist on this website)

  • and are essentially synonymous and they can both co-exist (given that both expressions are common)

It is possible there are a lot more duplicate tags.

Tags that need improvement

There are tags whose definition is not clear or inconsistent with the existing posts on this website. For example, the tag , whose current definition states

For questions about chat-bots. NOT for questions about how to program a chat-bot, as those kinds of questions are off-topic.

However, one of the highest upvoted question on this website with this tag is What are the latest methods to train a chat bot?.

Other tags that possibly needs improvement are and . In their current description, they have no relation to AI (so they should not even exist). These tags could be used for questions that are both related to biology (and the brain) and AI at the same time (e.g. human-inspired AI, whose tag, btw, already exists on this website).

I suspect that there are a lot more tags that need an improvement, in terms of description and scope.

| |

For investigating the tag problem in detail, a look into the tag section make sense. Right now, the list contains of 15! pages with over 500 tags in total. According to the Stackoverflow help section it's possible to delete a tag. For doing so, the tag must removed from all questions, and then the system will delete the tag at midnight with a script.

A possible way in dealing the tag problem is to scroll to the end of the list and search for seldom used tag. They can be unlinked from the question. In the first step, the list of all tags can be reduced to under 100 tags in total, and then it make sense to discuss which of them can be renamed and aggregated into groups.

Solving the tag problem with more tags?

A naive approach in dealing with the tag issue is to simply add more tags to the list. In the hope, that this will classify the questions better. The mentioned Stackoverflow website contains of 50000 tags. But the amount of questions in the system is greater. The ratio is 0.0028 tags per question. In contrast the ratio at SE.AI right now is 0.1 tag per question.

| |
  • $\begingroup$ You're right that if we disassociate a tag with all questions it is currently associated with, it will be deleted. I've done this in the past with tags I found unnecessary. $\endgroup$ – nbro Oct 22 '19 at 19:37
  • $\begingroup$ I am unsure whether there is a way for us to mark two tags as synonyms. I think that doing so is preferable to deleting tags. If we delete tags, users are likely to try to re-create them in the future. The tags exist because users (who likely did not know how to express what they meant) thought it was a good idea to create them. If we cannot link tags as synonyms, our moderation load would be greatly increased by trying to delete poor tags, especially as the tag counts increased. $\endgroup$ – John Doucette Oct 31 '19 at 1:42

It's telling that it's much easier to create tags than remove them, and I have to wonder if this is intentional or just a fail-safe, to avoid untagged questions.

I like tags in general because they allow degrees of specificity. For instance, most of our questions involve machine learning, but in relation to what?

That said, there is a lot to digest in your post and I am still going through it, so I don't have any specific response at this time, other than to note that Bayes is a good example.

| |
  • $\begingroup$ What does Bayes (the person) have to do with AI? It is not directly related to AI. If you put yourself in the position of the asker, it is highly unlikely that someone will ask a question about the person Bayes, which would be off-topic on this website, by the way. Even Math SE does not have such a tag. I've been a regular user of many SE websites, not just this one. All other websites do not create tags for every possible topic, especially for remotely related topics. $\endgroup$ – nbro Oct 26 '19 at 2:49
  • $\begingroup$ I actually think that tags associated with specific AI topics can surely be created. For example, ant-colony is a tag associated with a specific sub-field of swarm-intelligence. Both of these tags are directly related to AI. These topics are studied by AI people. On the other hand, there's no AI field that studies the person Bayes. We only use Bayes' work. $\endgroup$ – nbro Oct 26 '19 at 2:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .