I mean either a tag is assigned to some information or not.

Lets assume i have a photo showing horses in a forrest. The horses are in the foreground and huge.

Why can't i tag this photo as "#animals" and "#nature" with a 100% animals but only a 50% nature.
Until we have fine working AI this might help a bit in relevance searching.

Is there any information manager that can do this?
I haven't seen it used anywhere. It just came up my mind while programming on my own pet project.

I think that is a great idea!


But then, I don't write software.

Hmm, you could have the software make venn diagrams showing the overlap between tags. Not all "horse" would be in "nature".

YIKES!

I think that is a great idea!


But then, I don't write software.

Hmm, you could have the software make venn diagrams showing the overlap between tags. Not all "horse" would be in "nature".

YIKES!

I've never seen anything like this, but I have seen two uses of tags that might be relevant: The first is "tag clouds" which show tags by how often they appear. There are WordPress plugins to do this for post tags. Although it still remains binary for individual posts, the aggregate data does therefore give greater strength to those tags which appear most frequently. The second is the use of "sub tags" in apps like Bear where you could do something like: #primary/horse #secondary/nature. I find that the folks at Bear have implemented sub tags in a very intelligent way that makes me wish other apps did the same.

I would imagine there is no standard for "tag" or "tagging". There are some ancient RFCs on the topic, but they are somewhat technical (e.g., how to handle tags in UNICODE).

Developers have different metaphors for tagging: "tags are groups" -- DEVONthink and Apple -- "tags are labels", etc. But always @Lothar is right, "tags" are binary.

I think what you're looking for might be a concept for "attributes" or "properties": a {label, value} pair. This is the basis of XML, Apple's .plists, and so on. The "label" is static, the "value" can vary. So you could have {animals, 100} and {nature, 50}, and so on.

Tinderbox uses attribute data structures of this sort. Some mind-mapping apps do as well.

... and of course databases are structured with attributes. A nice personal solution for making what you seem to be looking for is Airtable.

I was only thinking about the user experience.

At the moment i consider tagging information as a an already learned task which is performed by people. I guess we technical guys often total underestimate the resistance of people to do more complicated things. So i think adding extra attributes will take way to long that it becames a habit.

In my idea the tag would be shown with anot just as a colored rectangle with the tag name but also a 5 star rating or a few pixel width tiny slider below the tag text where you can immediately change the value.

Special thanks for the the subtag idea. This is definitely something i will look into because i haven't seen it yet. But i don't think it does what i try to do with non binary tags.

Lothar Scholz wrote:

I was only thinking about the user experience.

At the moment i consider tagging information as a an already learned
task which is performed by people. I guess we technical guys often total
underestimate the resistance of people to do more complicated things. So
i think adding extra attributes will take way to long that it becames a
habit.

There appears to be a major trend in the direction of machine learning and neural networks for identifying and classifying "stuff." For example, if you use Google Photos, you can search for "dogs" and it will return your photos which contain dogs. As this technology improves, the user's own tagging will probably be reserved for esoteric or specialized vocabulary. Of course, this technology also requires users to give up their privacy rights to their data in order to help train the software; but the number of people who will not consent to this is likely to remain a small minority.

Paul Korm wrote:

I think what you're looking for might be a concept for "attributes" or
"properties": a {label, value} pair. This is the basis of XML, Apple's
.plists, and so on. The "label" is static, the "value" can vary. So
you could have {animals, 100} and {nature, 50}, and so on.

Tinderbox uses attribute data structures of this sort. Some
mind-mapping apps do as well.

You can sort of do this in ConnectedText as well. There are attributes and properties and you can build your own tagging system (e.g. by having a drop-down combo box with a range of values), with associated queries. But it requires a somewhat advanced familiarity with CT, and it does require some effort to set it up. You can also have sub-tags (sub-categories) of a tag that can be displayed as a tree hierarchy.