Thoughts on Tags

[Brad]

I've come to the conclusion that normal tags (as in the words you attach to pictures and the like to make them easier to search for later) are not enough. Using only tags doing some things which you would think would be simple are in fact annoyingly difficult. For example... say you have a database of photos and you want to get a list of all the people who attended some event. With the standard tag method you'd tag the event name and one tag each for every person in the photo. Suppose however that somebody also decided to tag some non-people data too. Like maybe the type of photo (portrait, macro, wide-angle,etc), random objects in the room (wedding cake, car, pizza, bowie knife,etc), and more. How are you supposed to be able to return a list of all of the "people" tags associated with that event? With normal tags it's pretty freaking difficult.

Tags need types. At least for people like me who like to tag the crap out of everything. If you could assign types to tags all of this would be much easier. It would make browsing soo much easier. Rather than looking through a giant, annoying, nearly useless tag cloud you could actually browse the list of people or the list of places or events or objects or whatever in your database.

This would complicate the UI for adding tags of course. No more single text box with a comma delimited list. You'd have to add tags individually and have a list of different types to choose from. But with that minor increase in inconvience when adding tags you'd gain a lot more power when searching tags. And it isn't like this would seriously complicate a tagging database. The tags themselves now just given a single bit of meta-data (crap I just thought about what it would be like if you could assign tags to tags... sheez.)

There are no major photo sharing/storing sites/software out there which does this (that I am aware of.) Not Flickr, not picasa, and not Facebook. I have seen it at least once before on some foreign language image board I stumbled upon though.

I'm not sure I want to be restricted to web based though. In fact... I'm not sure I want to be restricted to tagging just photos. Any type of file which you have a large collection could benefit from a tagging database. I'm thinking like a library of sound effects used in game development.

Also... what about photos that you've backed up and maybe you don't actually want/need always to be accessible on your current computer. When using an offline storage method like usb hard drives, dvds, blu ray discs, usb sticks, wax cylinders,crystals etched by red laser beams... whatever the storage method it would be great to be able to take the tag database with you.

That brings up another point... the tag (really meta-data) database needs to be lightweight. I'm looking at you Picasa... This shouldn't be a problem until you start thinking about storing thumbnails in the database. Then you need to make a bunch of decisions about the size and quality of these thumbnails versus your desire for a small database.

All of these design choices however seem to make me the enemy of sharing. Because if you make a database that can go along with the pictures on a USB stick it isn't exactly central or accessible by other users. The only real way I see to remedy this is to use some sort of distributed synchronization approach. You'd have to create a modified version of the database program (web-based of course) where the other non web-versions of the program could upload the pictures to and synchronize the meta-data. Thus giving you local and remote copy of the photos both with the same meta-data. This also helps if the photo-server goes down in flames; you've still got your local copies with their meta-data. Which photos get uploaded to the centralized server would of course be up to the individual users. Also the central server would be able to restrict file types and sizes and require login authentication.

Anyway that's my rant. If you were able to suffer all the way through it then let me know what you think.

[Nick]

A crude version of tag categories can be done in drupal already. One just needs to setup three taxonomies used for tagging things; a people category, a location category and another for interesting items in the photo or for its main subject (hats, waterfalls etc.)

I will add this to the drupal 7 test site (now reinstalled yet again)

[zourtney]

You just wrote the thesis for my college senior project Basically: yes to all. My solution was a tag hierarchy, but in retrospect, that's way too rigid most of the time. You kind of need an "is a" field and a "related to" / "carry forward" type of field.In my experience though, no one cares. Not enough to actually tag stuff anyway. Even if I started tagging everything today, and it had some awesome, usable UI, it'd still suck up hours of my time. I'd rather be sleeping, taking more pictures, or something (I still have 80,000+ unrecognized faces in Picasa...). And my tagging method changes over time, so old tags almost become less relevant.Facial recognition seems to work because it can be automated. And because is pretty static (that's still Frank, he's still a person).So yeah. A picture's worth a thousand words. But they might not be tags. There's got to be a better way...

[Brad]

I think you are right, what I am proposing is probably overkill. At least for everyday pictures anyway. If I wanted to create a stock photo library of photos which I plan to use again for whatever reason (other than just looking at them for nostalgia) I think it would be useful.

Actually anything related to assets could benefit from this. Given a sufficient number of assets it might be worth while to actually tag them. I think the real key is planned reuse versus required tagging work. If you plan to resuse the asset enough so that the cost of tagging it is less than the benefit of being able to find it easily again then it is worth tagging.

Most of the pictures we take are almost take and forget type pictures. You take the picture download/upload it to whever to show people and then you forget about it. Until such time as you decide to look through all your pictures again. So the value of tagging every single one of those pictures is much lower than the amount of work necessary to do all the tagging.

The expected amount of reuse for assets is much higher. If you've got a library of stock photos, illustrations, sound effects, music, 3d models, design sketches, etc... you are going to want to be able to search through that effectively.

Music libraries already basically already work this way. Rather than using tabs you're given a few metadata fields to assign values to (song title, album, artist, genre, year, etc...) This is basically like having a fixed set of typed tags. I mean tags with types is basically the dynamic version of this. People take the time to set all the metadata on their mp3s due to how often they reuse them. Which is much more often than their photos.

So, what I really want is a tag database which is not restricted to photos and does not require a web server with a lightweight database to use on any sort of collection of files I plan to reuse or search through often.

Nick: I agree with Courtney that picture galleries are not reused enough to warrant the amount of work necessary for complicated tagging (and maybe not regular tagging). It would still be an interesting experiment to see what crazyness you can do in drupal though.

[zourtney]

If you really did want to hack together a web-centric version (and wanted to get your coding fingers going), Django may be a better platform. But...if not, I'm not really sure what to use. Maybe that mythical database driven file system Microsoft was working on a decade or so ago? :PAll this reminds me that I need to try and restore my Picasa database....I backed it up when moving to Win8, but haven't touched it since...