Mark Boulton

24 Responses to “Tags and Folksonomy — I need your help”

1. 
   SteveC said on: July 25th, 2005 at 11:35 am

   Hmm, I often use tags to nav­ig­ate through posts on some very large sites, albeit, I don’t often find them any more use­ful than cat­egor­ies. Except on the sites you men­tioned. The likes of flickr, deli­cious and wists are really made bet­ter by tagging.

2. 
   Gerard McGarry said on: July 25th, 2005 at 12:47 pm

   Mark — I’m a cat­egory freak (prob­ably was a lib­rar­ian in a former life), but I don’t get much use out of tags. In flickr, they’re handy, but I wouldn’t miss the tags feature. 

   In del.icio.us I think it’s almost too easy to cre­ate new tags for your­self and you run the risk of going over­board. You could waste an awful lot of time micro-managing your link tags!

3. 
   Kevin Evans said on: July 25th, 2005 at 12:47 pm

   tags! It’s some­thing we’ve also been wrest­ling with here at Glam­or­gan. So far we seem to be using them purely on the CMS side of things. For example, events tag cloud is an admin side view of our tag­ging of ‘events’. For the user side we tend to replace them with ‘expan­ded inform­a­tion’, e.g.  A page from our site, below the fold we provide a “you may also be inter­ested in ..” sec­tion, which is tag driven.

   For pure tag clouds we so far are only exper­i­ment­ing with it on our soon to be released blog portal, and there are cer­tainly some reser­va­tions within the team.

4. 
   Ryan Nichols said on: July 25th, 2005 at 5:46 pm

   I think the blog post (the link from Kevin) makes a solid point about tag­ging. It’s a dif­fi­cult concept when your tag­ging text. 

   I think in many situ­ations using solid user research to develop a tax­onomy is much more effi­cient than rely­ing on user sup­plied meta data about the content. 

   But it depends of course on the nature of the con­tent. If it’s 100% user sup­plied then it can have more weight…after all you have no idea what the con­tent IS.

5. 
   James Archer said on: July 25th, 2005 at 7:14 pm

   I think tags are a good idea and we’ve seen some very good examples of how they can be used well, but all the hul­laba­loo about them far out­weighs their actual usefulness. 

   On Flickr, for example, the reason tags work is because there are so many people doing it that some generally-useful pat­terns emerge, and because the indi­vidual con­tent units aren’t par­tic­u­larly import­ant by them­selves.  On a smal­ler site (e.g., a cor­por­ate intranet), that wouldn’t neces­sar­ily be the case, because much-needed doc­u­ments might be poorly tagged by a well-meaning employee (e.g., “tpsre­ports” instead of “tps”). 

   On pro­jects like that where it’s import­ant that indi­vidual con­tent pieces not be lost, and where the audi­ence and con­tent quant­ity isn’t large enough for pat­terns to emerge des­pite widely-varying meth­od­o­lo­gies, a more formal approach would prob­ably be more appropriate. 

   It’s hard to give more spe­cific recom­mend­a­tions without know­ing more about what you’re doing, but I’d strongly recom­mend that you check out faceted clas­si­fic­a­tion as an option.

6. 
   Mark Boulton said on: July 25th, 2005 at 9:47 pm

   James — You’re right about tags being depend­ent on a large group of people sub­mit­ting, which in turn gen­er­ates pat­terns. The pro­jects I’m deal­ing with wouldn’t neces­sar­ily have this turnover of content.

   I’ll give faceted clas­si­fic­a­tion a look.

7. 
   Bill Lazar said on: July 26th, 2005 at 12:39 am

   Two things to consider. 

   Who is doing the tag­ging? If the con­tent author does it then the folks aspect of folk­sonomy is mostly lost and James’ point about quant­ity is magnified. 

   Are the tags used as the only sub­sequent means of access (as is cur­rently the case with deli­cious) or does the sys­tem sup­port search as well (Yahoo MyWeb, RawSugar–where I work)?

8. 
   J R Sinclair said on: July 26th, 2005 at 12:55 am

   I’m a PhD stu­dent cur­rently study­ing this whole, messy, area of user-centred cat­egor­isa­tion. I think James Archer summed it up well. Folk­so­nom­ies work when lots of people find that tag­ging bene­fits them each indi­vidu­ally. Then you can get interesting/useful inform­a­tion from the sheer volume of people tag­ging for their own bene­fit. You can then do won­der­ful things with stat­ist­ics and prob­ab­il­it­ies and whatnot. 

   How­ever, with cor­por­ate intranets and know­ledge man­age­ment sys­tems, etc., you often don’t have enough users to make free-tagging use­ful. You could how­ever, allow tag­ging from a restric­ted vocab­u­lary to pre­vent the synonym-tagging prob­lem (eg. what one per­son calls ‘tps’ another calls ‘tpsre­port’). The tricky thing is get­ting this restric­ted vocab­u­lary to match the vocab­u­lary that the users actu­ally use, rather than what the designer thinks the users use, or even what the man­agers think the users use. Some writers sug­gest that you allow free-tagging ini­tially, then come back and ana­lyse the data later, and con­sol­id­ate the vocab­u­lary into a finite list of allow­able tags/categories. This may not be prac­tical in all situ­ations however.

9. 
   Allan White said on: July 26th, 2005 at 1:13 am

   Bill’s point is good: who’s doing the tag­ging? One thing that annoys me about folk­sonomy sys­tems is when I’m not presen­ted (as a user) with a list of tags to choose from (I must come up with my own). I think someone men­tioned that above. 

   I’m think­ing about using tags for my next revi­sion of my port­fo­lio. This is because many of my pro­jects have mul­tiple media types or roles. BUT, I’m also going to apply cat­egor­ies to the “entries” — so con­ceiv­ably users could nav­ig­ate that way, as well. There’s also the search engine to con­sider — dif­fer­ent users browse differently. 

   I agree that there should always be mul­tiple paths to the information.

10. 
    Jeff Barry said on: July 26th, 2005 at 5:56 am

    The lib­rar­ian within me (and I am actu­ally a lib­rar­ian) wants to say that tag­ging will never be as use­ful as a con­trolled vocab­u­lary. Yet, I think that user gen­er­ated descriptors will evolve thru com­munit­ies such as tech­nor­ati and Flickr as these sys­tems put more thought into this area. The tag-cloud is a good visu­al­iz­a­tion tech­nique. I think that there are going to be some big devel­op­ments in the UI for tag­ging in the com­ing year. And I must say that I’m noti­cing more new traffic on my blogs com­ing via tags than thru search engines. There’s an inter­est­ing dis­cus­sion on tag­ging tak­ing place at tagsonomy.com

11. 
    Bill Lazar said on: July 26th, 2005 at 3:51 pm

    Con­sid­er­ing how early we are in the use of tags there’s no doubt that bet­ter UIs will be found to express them and bet­ter ways to determ­ine what tags to show at given points in a search path. That’s actu­ally some­thing our com­pany is work­ing hard on at the moment. I think clouds are inter­est­ing but when data­bases get large enough to cover a mean­ing­ful por­tion of the web the quant­ity of tags may over­whelm their usefulness.

12. 
    Tim Parkin said on: July 27th, 2005 at 10:02 am

    My main prob­lem with tag­ging is that it ‘gives up’ the attempt to provide a uni­ver­sally under­stood gloss­ary. Mul­tiple dif­fer­ent ways to cat­egor­ise the same ideas occur hence someone search­ing on just one of them will return a sub­set of data that doesn’t match what they want to find but matches posts by people who describe things in the same way. An upper onto­logy, how­ever innac­cur­ate, will at least provide a con­sist­ent (if flawed) way of cat­egor­ising inform­a­tion. Per­haps the tags can be col­lec­ted and assigned to a hier­arch­ical cat­egor­isa­tion struc­ture and syn­onym lookup inform­a­tion provided, but it would seem to me that get­ting a simple ini­tial cat­egory hier­archy, although more dif­fi­cult and con­tro­ver­sial, would provide bet­ter inform­a­tion clas­si­fic­a­tion as time pro­gresses (let the users have some influ­ence on the cat­egory hier­arch­ies / or faceted clas­si­fic­a­tion schemes maybe)

13. 
    Mark Boulton said on: July 27th, 2005 at 10:20 am

    I think you’re right Tim. My prob­lem with user tag­ging is the same as yours. Whilst tag­ging the inform­a­tion is all very well, provid­ing an inter­face (search, nav­ig­a­tion etc) to access the inform­a­tion is always going to be dif­fi­cult for the reas­ons you just stated. 

    One way to per­haps do this is to provide, in part, the tags avail­able to the user. That would all get very dif­fi­cult to man­age though, espe­cially when you’re talk­ing of hun­dreds of thou­sands of documents.

14. 
    Tim Parkin said on: July 27th, 2005 at 10:44 am

    have you looked at wikipae­dia tag­ging? Use the wikipae­dia page id’s as your tags… there are very strong defin­i­tions for what each ‘term’ means. For instance if I wanted to tag my writ­ings on the python lan­guage I would use Python_programming_language as my wikipae­dia tag (so it’s def­in­itely not pets or monty).

15. 
    Bill Lazar said on: July 27th, 2005 at 5:52 pm

    I think that the issue of usage will come out in the wash over time as sys­tem developers learn to use the data to find con­nec­tions. There will also be bet­ter ways of present­ing exist­ing tags so they can be reused rather than retyped. To some degree our sys­tem does both already though we’re not claim­ing any­thing mroe than a start in the right dir­ec­tion yet.

16. 
    Allan White said on: July 27th, 2005 at 11:16 pm

    Mark B.: Any thoughts of how one might apply folk­sonomy approaches to Expres­sion Engine? I’ve been check­ing out the Keywords plug-in, and I won­der if you could cre­ate a UI for pre-set tags (like Tim men­tioned above). 

    I agree that let­ting users tag how­ever they wish is likely to cause con­fu­sion for oth­ers searching.

17. 
    Bill Lazar said on: July 28th, 2005 at 12:22 am

    Allan, I’m not an Expres­sion­En­gine user but Word­Press (my pref­er­ence) has sev­eral tag­ging plu­gins avail­able and you might look at those for ideascraping.

18. 
    Mark Boulton said on: July 28th, 2005 at 8:55 pm

    Allan — Yeah, I’ve been think­ing about tag­ging for this site actu­ally but more from a keywords aspect. 

    What I’d like to do is turn my ‘of interest’ links into related links which can be linked to from journal entries. I’m a little bit frus­trated with the links turn­ing into an archive which is totally flat in terms of categorisation. 

    So, the thing now is to come up with a suit­able UI which shows the rela­tion­ships between the journal entries and the ‘of interest’ links.

    Oh, I’m using the Link­slist plu­gin for this which has built in sup­port for tags and cre­at­ing tag clouds. Couple that with the Keywords plu­gin and you have a com­plete sys­tem for hand­ling it. 

    The pain is migrat­ing all the data manually…

19. 
    Allan White said on: July 29th, 2005 at 5:08 pm

    So, the thing now is to come up with a suit­able UI which shows the rela­tion­ships between the journal entries and the ?of interest? links.

    That’s an inter­est­ing task; what might that look like? A list of tags at the end of a post (“design, typo­graphy, Macs”) that link to a link list cat­egory page? Do I under­stand your use case properly? 

    I (and oth­ers) are also curi­ous to see how you might solve this in an EE con­text. There’s a dearth of qual­ity tutori­als for that deep CMS, and one reason many from that com­munity come here!

20. 
    Allan White said on: July 29th, 2005 at 5:18 pm

    [J. Sin­clair] How­ever, with cor­por­ate intranets and know­ledge man­age­ment sys­tems, etc., you often don?t have enough users to make free-tagging use­ful. You could how­ever, allow tag­ging from a restric­ted vocab­u­lary to pre­vent the synonym-tagging prob­lem (eg. what one per­son calls ?tps? another calls ?tpsre­port?). The tricky thing is get­ting this restric­ted vocab­u­lary to match the vocab­u­lary that the users actu­ally use, rather than what the designer thinks the users use, or even what the man­agers think the users use. Some writers sug­gest that you allow free-tagging ini­tially, then come back and ana­lyse the data later, and con­sol­id­ate the vocab­u­lary into a finite list of allow­able tags/categories. This may not be prac­tical in all situ­ations however.

    Idea: what if you let users sub­mit a ‘tag request’ in this scen­ario? If you had a list that was ‘good enough’ for most uses, then let users sub­mit requests for the arcane ones (’tpsre­port’) to a human ‘tag cur­ator’ for consideration.

21. 
    Tim Parkin said on: August 1st, 2005 at 9:52 am

    Hi,

    Promp­ted by this dis­cus­sion, I’m try­ing a slightly dif­fer­ent approach based on a com­bin­a­tion of hier­archy and folk­sonomy (or plain metadata). Here are the steps I’m build­ing in : (quick gloss­ary — hcat = hier­arch­ical cat­egory, kw = folk­sonomy tag or keyword) 

    1) scan post for words that match hcats and show them with tick boxes for selec­tion
    
    2) let the user add their own kws and scan them for related hcats
    
    3) let the user tick appro­pri­ate hcats
    
    4) allow the user the option to (i) map any remain­ing kws to cat­egor­ies (syn­onyms) or (ii) place the kw in hier­archy (pos­sibly moderated)

    5) store the remain­ing kws in a ‘metadata’ field. 

    step 5 can be used if people later wish to add a new cat­egory and find related items (e.g. I add Corn­wall as a kw and a later users decides to cre­ate a real cat­egory for it and searches in meta data for it). 

    This may seem long win­ded but it allows people to just add keywords if they like but it helps them to add (and start to cre­ate) hier­arch­ical categories. 

    Any thoughts would be appreciated.

22. 
    Mark Boulton said on: August 6th, 2005 at 4:45 pm

    Hi Tim,

    That seems like an inter­est­ing solu­tion to a dif­fi­cult prob­lem. I’m not sure how the tick­ing would work within a body of text. One of my major gripes with includ­ing lots of links within text is it can hinder the reader’s eye when scan­ning the copy. 

    Be nice to see it in action.

23. 
    Tim Parkin said on: August 7th, 2005 at 11:12 am

    Thanks Mark,

    I did intend for the tick­ing to be done once the art­icle is fin­ished so the tick boxes would appear at the bot­tom of the text. This could be done whilst the art­icle was being writ­ten (but not effi­ciently with a large ontology). 

    Tim

24. 
    David S said on: August 15th, 2005 at 6:55 pm

    This is some­thing I am strugling with as well, and the com­ments many people have made already are helpful. 

    It sounds like this may be a bit basic for some (it goes ‘back’ as far as hier­arch­ies), but I found the fol­low­ing art­icle a good start: http://wiki.osafoundation.org/bin/view/Journal/HierarchyVersusFacetsVersusTags