Hierarchical tags
Tagging is better and more flexible than collections - but if you make possible creation of tag trees, you can combine best from two worlds. You can see one variant if WebResearch (http://www.macropool.com) - user can just select checkboxes for categories, and can see categories as collections. In free-tagging this can be done as "parent/child" notation.
To me, this approach seems more flexible, because it can be used with more than just tags. Or am is there some other advantage to/reason for tag trees that I'm not seeing?
You can also use them in negation with the "is not" operator to find all results that /don't/ appear in the other saved search, use multiple saved searches as conditions to combine results, set up dependent chains of saved searches, etc.
(Note that while a saved search doesn't let you choose itself as a source, it doesn't currently check the entire chain, so it's possible to set up a recursive loop (e.g. one search dependent on another that's dependent on the first). Mozilla will catch it and throw an error after a couple seconds, but it's still something to be avoided.)
Thanks, I didn't notice that... may come in handy :)
This has another advantage: if I use free tagging, it is hard to remember all tags, but if you show existing in the list, it wiilbe long and not so usable. Tree is much more compact and IS usable, from my experience. To avoid long typing in free tagging you can enter last tag, and application must suggest you possible variants - in my example it would be "IT/Programming/OO Programming/C++" (and, may be, ""IT/Programming/Languages/C++").
http://netapps.muohio.edu/blogs/darcusb/darcusb/archives/2006/09/09/zotero-and-the-practical-semantic-web
http://www.betaversion.org/~stefano/linotype/news/85/
I think from a UI perspective the flat approach with auto-completition is best. It should scale well to thousands of tags in fact.
But there may be value too in being able to organize the meaning of tags as well, and that requires something like Stefano's approach above.
The problem with all this is that tagging really needs to account for the fact that users want to use their own taxonomy. I don't mind including a few tags, and using search to achieve the rest. I do actually get a little bothered by the fact that Zotero imports third-party tags (from the LoC, NYT, etc.) by default, since their classification schemes (at least the labels) often conflict with mine.
Anyway, applications for information organization are rare and primitive now (most of them use simple tree for organization - it's shame), so Zotero is a good advance.
Of course, a simple modification to the way collections work could enable this as well: make it so references in a sub-collection show up in the parent collection as well. (Or it could be an option.) Then add a reference to the collection for C++, and it shows up in the Programming collection. (I found myself wishing it worked that way just yesterday.)
Then again, since references can be in multiple collections there's only a fine line between tags and collections.
OTOH the simplicity (and ease) of tags is one of their greatest powers.
We are touching here on a bigger issue of knowledge/information sharing and management. Simplicity is a prerequisite for an organisational system to be used by lots of people (and not just 'experts'), but rich semantics are needed for powerful functionality (like using automated reasoning in the vision of the semantic web).
I have a qualified suggestion and a question:
What other efforts at combining or dealing with this sort of thing are happening out there? Lots has been done with tags. eg: tag clouds, del.icio.us is one (last I looked it was just tagging, little (no?) semantics)...
Should Zotero be set up to allow plugins to manage tags? Those who want semantic richness or to reference some experts ontology could do that (if the plugin was written by someone :) Services may come to be (perhaps in del.icio.us??) that could be used to manage zotero tags too etc..
following stefano's ideas in the ling above, maybe tags could have extra info indicating where/how it was created (I guess the general UI may not want to display this info?).
FYI: There's been a few relevant comments here:
http://forums.zotero.org/discussion/7168?page=1#Item_16
and here
http://forums.zotero.org/discussion/700/equivalence-of-collections-and-tags/
and here
http://forums.zotero.org/discussion/4006/hierarchical-organization-of-tags/#Comment_17182
I'm no expert on tagging but my impression is that simple hierarchical and semantic strategies are proving the most popular and useful. Mainly: (1) defining a hierarchy by embedding delimiters in tags, and (2) so-called "triple tags" or "machine tags".
I'm sure there's tons of academic research too.
I guess I see two courses for zotero:
- try to keep track of tag format standards and tag management web service standards, and implement as seems appropriate
- engage or someone become an expert in tag management to guide/advise/make useful and informed systems
This discussion is part of the second point, but it seems none of us here claim that expertise. I'm pretty keen, but my phd is pretty all consuming at the moment, and will be for the next 6 months at least..
Hmm.. it strikes me that tag management services/standards are on the way. Zotero could be an interesting place to start that trend..
Dr. Ian Evicko
(;
(even if I'm not a PhDr yet :)
This becomes more critical if it ever becomes possible to do tagging of PDFs in a useful way at the level of sections of text, not just whole documents. I think I've given up on using Zotero as a PDF reading and markup tool, partly because of this issue, although I dearly wish it were possible. I've suggested before that tagging the PDF of "Moby Dick" with the tag "whale" doesn't do much good, although if it were possible to tag at the level of selected text, then using hierarchical tags would allow for a close analytical reading of a text. Even if it were possible to do this sort of tagging in an external PDF reader, if Zotero were then able to use those tags, that would be useful. So far none of the PDF reader programs seem to have useful ways to do any of this.
But again… unfortunately, a lot of opportunities are not visible enough in the documentation
If there is anyone interested, I will be pleased to provide the code.
So far I am however not sure how to do this best because - as you can imagine - this required not only some chances of the Zotero core. Currently I got everything running based on Zotero 3.0.7
http://groups.google.com/group/zotero-dev
or even put up a pull request to Zotero - I don't know how Dan sees this and it depends on the UI integration, but there is certainly no fundamental opposition to including this in regular Zotero.
1. Make collections more flexible, so that on a case-by-case basis the user can define whether a collection includes sub-collections, or (although that would also require adding conditional groups) refining higher collections. http://jabref.sourceforge.net/help/GroupsHelp.php
2. The other way, looking at tags, would be to allow the user to organise their tags themselves, but in a loosely defined way. So tags can still be 'flat', but users can create relationships between them, using concepts. So, I can have a concept "Projects" and can add tags that define particular projects... Not all-encompassing, but already quite useful, especially if a concept can contain both tags and concepts (which I don't think it can in Bibsonomy).
http://blog.bibsonomy.org/2010_04_01_archive.html (second entry)
So, in summary, strong support for hierarchical tags.
(I know I can already search by choosing a collection name from the drop-down menu. But when the number of categories is large, drop-down menus become cumbersome.)
For such jump not to interfere with quick search for titles and others, search for multiple collections would have to happen only when the user specifically chooses to search among collection names, until a better solution comes along. One idea is to have a dedicated search box for collection names like the one we have for tags.