Hierarchically-organized tags?

onedayonenoodle

I think it would be nice to have tags that are hierarchically organized. For example, you have tags "brain", "neocortex" and "primary motor area". Because "primary motor area" is always part of "neocortex", which is always part of "brain", it would make sense that all the items tagged with "primary motor area" are automatically tagged with "neocortex" and "brain" as well. Doing this manually for every item is extremely tedious, especially when you have 1000+ items. The tag selector box GUI needs to be drastically updated to support this functionality, but I believe that the situation above is not uncommon.

adamsmith

lots of threads on this already:
https://forums.zotero.org/search/?PostBackAction=Search&Keywords=hierarchical+tags
I think there's no general opposition, but it's nothing that will be implemented by core devs any time in the short or even medium-term, so if it's going to happen, it'd have to be a patch (which would be very welcome, but for such a major feature, also like a lot of work, including back & forth with devs about details of implementation.)

onedayonenoodle

Thanks. Sorry, I searched for "tags" only.

Glad to know that I wasn't the only one. It requires clearly a lot of work.

Seikou

I can't see the point of hierarchical tags, at least not in this case. Tagging should be exploratory and aim at making and remembering connections that are not self-evident, so it should emerge from your own research rather than reflecting the conventional structure of your discipline. If you know the topic of your research, you certainly know that "primary motor area" is part of the neocortex and the brain, so making that relation explicit to yourself is rather pointless. It's not like you have to help novices find what they're looking for, which is the very reason librarians must use hierarchical sorting. In your case, triple-tagging would satisfy your sense of logics, but would be anti-economic and wouldn't serve any practical purpose. If you really need to see that tree structure, use collections instead.

adamsmith

Tagging should be exploratory and aim at making and remembering connections that are not self-evident

no. The whole point of having tags, collections, saved searches etc. is that there is no "right" way to use any of them. If you use tags that way and it works for you that's great, but it needn't work that way for everyone and that's OK. We want Zotero to be flexible and able to accommodate different styles of organization.

Seikou

Well, I used to work at a help desk in a computer lab and I can tell you most programers can't even start imagining to what uses their programs can be put "in the wild". Actually that's what hacking is all about. Just think of newsreaders and the NTTP protocol, which ended up being abused and repurposed to share files. Yet, I've seen people writing their CV's on a spreadsheet only because they didn't know how to make proper tables and aligning in a word processor. Some tools are better meant for particular purposes, because their strengths and limitations dictate them. And anyway, in Zotero's case, developers have already voted on this for years now: no hierarchical tagging for you.

onedayonenoodle

"If you know the topic of your research, you certainly know that "primary motor area" is part of the neocortex and the brain, so making that relation explicit to yourself is rather pointless."

I do not agree with your point at all. I certainly do know the fact that "primary motor area" is part of "neocortex", and it doesn't make any difference, because it is not about to remind me of the fact. Rather, it is about how to effectively group your library items.

I am using double or triple tags, but these methods are prone to errors. Some of the items tagged with "primary motor area" are bound to miss the "neocortex" tag inadvertently. Then the "neocortex" tag becomes unreliable.

At the moment, you can manage hierarchical collections instead of tags, and I'm fond of that feature. Hierarchical tags will serve for the similar "Use Case" in another way.

aurimas

While onedayonenoodle sort of hints at this, the major point of hierarchical tags is not for you to see the parent tags when looking at an item, it's to be able to search by the parent tag and pull up the item with a more detailed (child) tag.

onedayonenoodle

Thanks, that's what I meant to say.

One of more complicated issues is whether to accept multiple parent tags at the same level for a child tag. It might work.

Seikou

onedayonenoodle, you actually agree with me because you found collections was the feature that better served your goals, but you now hit the crux of the problem. The very essence of tags is that they're not exclusive, you can assign as many as you want. However, if you assign one tag to multiple parents, you don't get a taxonomy that can be represented with a tree like with collections, you get a network or a cloud, which is what the tag selector tries to represent in a partial way. It's a difficult trade-off between easiness of representation and fidelity to actual relationships.
I don't think this can be solved with hierarchical tags. What I really miss in Zotero is "concepts", i.e., the possibility of freely grouping tags in a loose way. That doesn't mean concepts should be tags in themselves and be directly applicable to items, or that they be recursive and therefore allow hierarchical representations.

onedayonenoodle

Yes, I think there will be a trade-off too. How should Tag Selector represent multiple parents? Are we comfortable with seeing a tag "A" as standalone, and as a child of tag "B" and "C" at the same time in the Tag Selector Box, i.e. multiple appearance of the same tag "A"? Not sure yet. If it goes recursive, it could get pretty complicated. But this could be solved by a smart GUI design as I wrote above. For example, hovering on a parent tag activates a small pop up pane in which you can browse all the recursive hierarchical children of that parent. You may end up having 5 layers of panes or more to reach a child tag at the bottom level.

Seikou

If you get familiar with genealogical software like I did, you'll realize those GUI problems are actually insolvable. I found I descend three times from a couple that lived in the 17th century. The only accurate representation of my family tree would have to be a three-dimensional model and still it'd be a very challenging one to make. Just substitute concepts for family relationships and you'll quickly get into nightmare scenarios.

onedayonenoodle

I've never used any genealogical software but the closest I can think of is NeuroTree, which is an online service that allows you to see "family trees" of researchers in neuroscience. It has very simple "tree" view, and it works fine. Perhaps, "marriage" is making the things complicated?
http://neurotree.org/neurotree/tree.php?pid=-1

I'm not saying the intuitive implementation is easy, but I'm still curious.

DWL-SDCA

This seems to be a discussion about automating a hierarchical taxonomy of tags similar to terms in a database thesaurus. I think that is a great idea. Open source thesaurus development software exists that could help with setting up the hierarchical relationships among the tags.

onedayonenoodle

I'm not sure exactly what kind of thesaurus software you're talking about. Could you paste a link to a screen shot or something? If you're talking about something like Visual Thesaurus, which shows you a cloud of linked words, that wasn't really what was thinking about.
http://www.visualthesaurus.com/

Seikou

onedayonenoodle: I just checked NeuroTree and found it rather off-putting. This is precisely the kind of oversimplification that should be avoided: do scholars truly draw all their inspiration exclusively from one mentor? It may satisfy some people's elitism and reflect the narrow burocratic constraints of academic life, but I don't think minds reproduce clonically like unicellular organisms, they rather "have sex" and fertilize each other. So it happens with concepts too.

DWL-SDCA: Thesaurus software for common language has little relevance in this case. Everybody believing scientific terminology and common language have anything in common should re-read Foucault's "The Order of Things", Eco's "The Search for the Perfect Language" and first of all Borges' story about John Wilkins' philosophical language and the Chinese encyclopedia.
Besides, if tagging were automatic and based on common language, it would only reflect common sense ideologies and be the exact opposite of research. My own research, for example, focusses on Hawaii. Should I file the tag "Hawaii" under "USA" or under "Polynesia"? Each option has different ideological underpinnings. Thus, I strongly believe all researchers should create their own tags in the process of getting to undertand their topic.

DWL-SDCA

I was not writing about some variant of Roget's Thesaurus but a database thesaurus. A modern structured database thesaurus is not based on common language. When you cite Foucault or Eco you indicate to me that we are arguing from the same point.

See the Medline/PubMed MeSH, the NAL Thesaurus for AGRICOLA, the Ei Thesaurus for Compendix, the PsycINFO thesaurus, etc.

The modern thesaurus need not have only a one-to-one upward correspondence where each single term can have only one broader term. The MeSH system is a good example of this. Thus, to your question about Hawaii the answer is "yes" and also perhaps Volcanic Islands.

I don't think that this discussion is about Zotero imposing terms or hierarchies but facilitating the process that can allow a user to structure their own system. There again we strongly agree, at least for individual Zotero users. Having agreed-upon tag structure for those in groups is a different matter.

I'm not suggesting that a thesaurus-like system be incorporated into the main Zotero program because I believe that few would need it or use it. This seems a candidate for a plug-in.

I use TemaTres as the open access base for my system's software and it's engine could certainly be adapted for other use as well. There are other thesaurus products.

onedayonenoodle

#Seikou

> I just checked NeuroTree and found it rather off-putting. This is precisely the kind of oversimplification that should be avoided: do scholars truly draw all their inspiration exclusively from one mentor?

Sorry, but simply the example I gave was not good. I didn't realize "pid=1" randomly picks up a person from the database (it's not reproducible, so we were talking about different neuroscientists). Have a look at the tree of David Hubel, who had multiple mentors and multiple (a lot of) students/postdocs.

David Hubel
http://neurotree.org/neurotree/tree.php?pid=112

> Besides, if tagging were automatic and based on common language, it would only reflect common sense ideologies and be the exact opposite of research.

I agree with this point.

Seikou

OK, after checking NeuroTree and MeSH, I realized why we're not understanding each other. The key question is: do we want Zotero to be a cataloguing database for huge, diverse libraries?

If in the affirmative, we certainly need a hierarchical classification, and preferably one that allows for placing the same item at different locations of multiple, intersecting taxonomies, like MeSH, or a "faceted" classification like the Universal Decimal Classification, which can qualify any node in the main taxonomy with "auxiliary tables" for language, geographical location, date, etc. In fact, for this purpose the best solution would be a wholesale import of those very same classification systems. They already provide ready-made controlled, standardized vocabularies that are useful in themselves, whether they be taxonomies or mere lists of allowed descriptors.

Unfortunately, when I have Zotero import tags from library databases, it only retrieves extremely idiosincratic terms that are completely useless to me. Maybe we need a plugin to translate standard library codes into tags? Also, it is not clear to me how Zotero could reconcile different "languages" like MeSH, PsycINFO, and UDC; some users only need the thesaurus for their specific disciplines, but others doing interdisciplinary research like me could only use a more comprehensive schema like UDC. Also, there are conflicting general classifications like UDC, DDC, LCC... This challenge is similar to importing tags in different human languages like English and Spanish. The only alternative would be painstakingly recreating your preferred schema by yourself. Nothing would spare you that toiling labor, and in this case hierarchical taxonomies are at a disadvantage because they require more work for recreating all necessary upper nodes.

My own take in this issue is: I'm not compiling a general library (not even for my own field) but only for my current research project, I don't need a comprehensive classification system, and I don't think Zotero will ever be good at that.

valentijn

I completely agree with DWL-SDCA (David Lawrence) that a plug-in offering functionalilty to define or import tag hierarchies (thesauri) would greatly enhance the power of Zotero, especially in managing large libraries.

The possibility to formulate queries from a command line, using complex boolean operators and more, would ideally form part of such a plug-in. The possibility to use nested saved searches is not an option for any serious work.

What you want, basically, is to be able to add standardized subject indices from a thesaurus to your documents just like databases like Pubmed do, and then approach them with a sufficiently expressive query language.

Two things are important: firstly, to be able to build thesauri from existing ones like MeSH, but tailoring (restricting) them to your own needs. Secondly, merging them, if your research covers a wider area than any of the available thesauri.

Do you have any idea, David, if and how this sort of thing could be achieved with little or no programming effort? Perhaps by combining Zotero with other programs out there? I looked at Tematres (thanks for the tip) and some dreadfully expensive commercial thesaurus editors. I understand there are a number of query languages, like SPARQL and RESTful with which to approach the resulting SKOS-output. But both my time and programming skills are limited, so building this dream-plug-in is not something I could take upon me, I'm afraid. But I was thinking, perhaps it would be posssible to access Zotero items through links from some other, more powerful program, and store the subject indices there? Do you perhaps have any suggestions?

Nosher

A hierarchical organisation of citations should be targetted at the final document you wish to produce, e.g. essay, report, or Wikipedia article, rather than just assembling your collected researches (- you'll probably need a network for that). Information in the final document produced will be in a linear form so a hierarchical organisation should be adequate. As has been said before if you wish your citations to appear under more than one parent grouping then sub-collections with duplicate entries are the way forward.

As I see it any hierarchical tag system should behave very much like the Bookmarks manager in a browser such as Firefox. In particular it should have the ability to:
* sort the contents of a group into alphabetic order
* use drag and drop to place groups and citations in any arbitrary order
* use drag and drop to move groups and citations to any location in the tree structure.

It should still be possible to view Zotero bookmarks alphabetically in the way you can now. As the Zotero screen is crowded the hierarchical tag system should only be visible as a pop-up window. It should also be possible to copy 1. the tags and 2. the hierarchical tree of tags to the clipboard for further processing.

It is already possible to do quite a lot of hierarchical organisation with Zotero as it stands. The multiple levels within a tag system can be separated by a vertical bar character, e.g.

activities
construction|materials
construction|dates
lighthouse|light|range

This method has the advantage that adjacent items on the 'hierarchical tree' are grouped next to each other. The downside is that tags are liable to be on the long side however users just have to be concise and try to limit themselves to one word at each level and abbreviate.

It has been found to be a good idea to maintain a list of the hierarchy of tags in a separate document so as to find your way around the system of tags more easily. This list can with rearrangement be used as the basis of your target document.