Lee Iverson

Thomas Vander Wal on Folksonomy

2006-03-29T18:43:44Z

In Online Information Folksonomy Presentation Posted :: Personal InfoCloud Thomas Vander Wal writes about tagging, folksonomy and taxonomy and rightly points out that they are not necessarily opposed or incompatible concepts. I'd go further than him though to suggest that they are, in fact, quite compatible.

]]> So let's start out by reiterating Vander Wal's terminology. He seems to be making the distinction between "tagging" and "folksonomy" by pointing out that tagging has been going on for a long time and that folksonomy is new, moreover that old style tagging "provided no certainty as to the vocabulary being correct in the perspective of all trying to use it." The distinction he seems to be making is between "creator-driven keywords" (tags) and something new represented by del.icio.us.

The something new is the triple: person, tag name, object. Anyone with a database background and a few minutes on del.icio.us realizes that what is going on is that there's a table for identities, one for tags and one for URLs and in the middle is an association table. The significant thing about this architecture (and the thing that seems to get Thomas to say "folksonomy") is the "tags" association table at the middle of this and its consequences. For one, whenever an individual tags an object, an entry in the association table is made for every one of the "tags" used (e.g. if user "leei" tags "http://www.ece.ubc.ca/~leei" with the words "me" and "leei", we'll get two entries of the form (leei,"me",URL) and (leei,"leei",URL) in the "tags" table. A user can add as many tags as he wants to, but all of them will be referenced to his user table entry. Similarly, an entry will be added to the "user_urls" table with the user's title and description of every URL he uses.

One of the primary consequences of the relational model and the database backend here is that queries can be made from any side of this association:

From a user, I can see all tag names and urls;
From a URL, I can see all users and tag_names;
From a tag name, I can see all users and URLs;
From a user and tag name, I can see all URLS;
From a user and URL, I can see all tag names; and
From a tag name and URL, I can see all users.

And, of course, I can obtain counts of the size of all of these sets and use those statistics to generate things like tag clouds etc. It is this attributed tagging in a public space that Thomas Vander Wal seems to prefer the term "folksonomy" for (hopefully, he'll correct me if I've got this wrong).

Now, that said, what is the relationship between this and taxonomy. Here I disagree with Vander Wal. Taxonomy is a way of managing an organizational system as a hierarchy (and he rightly points out that many taxonomies may exist within the same domain), whereas the tagging or "folksonomy" models now in place seem to completely ignore the issue of hierarchy. This is clearly a deliberate choice (see Clay Shirky for example) but is it a necessary one? I'd suggest that it is not. The fundamental nature of folksonomy is a personal organization of objects of interest using simple markers (tag names). I do not believe that this necessarily implies that there can be no expressed relations between those markers. Some of the markers may express categories in the user's mind and there may be relationships between those categories (e.g. sub- or super-category relations or mutual exclusion). As with all technologies, it is important to distinguish between the manifestations of a technology and its fundamental nature.

So, I agree that taxonomy and folksonomy are not incompatible. In fact, I believe that taxonomy can be a part of folksonomy if the user finds it useful. The oppositions here are clear, they are between authoritative or prescribed vocabularies and personal or socially constructed vocabularies and between structured, categorical organization systems and flat or unstructured organizational systems. The table below shows these two axes with representative examples of the kinds identified by these properties. Of course, these axes are by no means binary, but more or less continuous. For example, one can be completely structured as in closed database schemas or semi-structured as in various RDF models that respect certain categorical relationships but leave definitions open and relationships largely unspecified such as Dublin Core.

Axes of organizational vocabularies.
	Structured	Unstructured
Authoritative	Domain Taxonomy	Glossary
Personal/Opportunistic	Hierarchical Filesystem	Folksonomy

Of course, the above is just a starting point, a sketch. This is really much more multi-dimensional than that. There is the personal/social axis, the private/public axis, the authoritative/opportunistic axis, and a few others. I think this is ripe for exploitation. I just don't think personal, tag-driven, opportunistic organizational systems are necessarily unstructured.

2006-03-17T00:09:27Z

In Lowering Barriers to Participation, Bradley Horowitz points out how tagging lowers the barriers to participation and thus allows ordinary users to create useful metadata. I think he's right on here. Especially on a system such as del.icio.us, the ease of use and triviality of creating tags allows pretty much anyone to create "personally useful" metadata. As an aide memoire, that is exactly what it should allow.

Is it authoritative? Is it useful for others? That's for someone else to decide. It's easy and useful for me, and if there are natural concordances with others, I can take advantage of those. That should be enough, and is.

My Favourite Cafe: The Beanery

2006-03-16T06:01:21Z

My favourite little cafe is a place called "The Beanery" on the UBC campus in the middle of the Fairview residences. Sonny and Sonya, who run the place, are two wonderful friendly hosts, the atmosphere is great, and UBC wireless is usually available. It's my second office.

It's located inside the Fairview residences. The entrance to the residences is at the intersection of Fairview Ave. and Western Parkway. There's a sign there for the Beanery, but it is actually all the way in the back of the residence complex, near Pearkes Lane.

EXTENT

How to Write a Proposal

2006-02-25T00:01:42Z

I've been asked repeatedly by students and others for advice on writing a proposal document. As with all documents, I believe that the best way to go about this is to concentrate on two things: the audience and the goal of the document. In general, a proposal is written to someone who has the ability to allow you to do something (i.e. money, resources, supervision) and your goal is to convince them to enable or allow you to do it. With that in mind, we can come up with a few guidelines.

]]> In essence, you are trying to provide a good answer to a simple question (from the point of view of your audience): should I commit my resources to helping this person accomplish their goal? Your goal is to get them to answer "Yes!"

One approach that I really like is to concentrate on the NABC's of your problem: the need, approach, benefits and competition. If you can present an argument that accomplishes the goals described by these principles, then you're well on your way to convincing your audience.

Well, clearly you start out with some problem you are planning to solve - that has to frame everything. So So , how do these NABC's relate to the problem, and how do they answer the question for your audience? Well, I find it useful to again, think about what questions you are answering for each topic.

Need: Why do I need to solve the problem? Who has the problem? How important is it to them? How much do they need it?
Approach:How am I going to solve the problem? How does this approach relate to the need and the people who need it? Why will this approach solve it?
Benefits:What will the benefits of this specific approach be to the users? How will it change what they are doing? How will the specifics of the approach affect these users? What are possible downsides and how can they be mitigated?
Competition:Who else has tried to solve the problem? How did they do it and how successful were they? How does your solution compare to theirs in terms of benefits and drawbacks?

In general, if you get this far, then you are well on your way to making a case. If you can't answer all of these questions, then you have more work to do. Now, that said, there are definite differences in style and specifics with different audiences.

For academic research the order is usually a bit different: NCAB. You need to further justify the need and make the transition to your approach by doing a review of the work that has already been done that: makes the case for your need, either directly informs or relates to your approach, and provides a baseline to compare yourself against.

For business proposals (and the NABC is sometimes referred to as a mini-business plan), the reason you end with a description of the competition is that that also helps to answer the money questions. How big is the market for this? What impact will it have on an organization in terms of cost-savings, market expansion, or increased productivity? How is the competition selling themselves and how does that relate?

Now, as to writing style and length, that's up to you. You have to make the case. But you should be able to make it at some level in one page, and also be able to produce a more detailed proposal in 20-30 pages. In the ECE department, we limit our Ph.D. proposals to a maximum of 30 pages. If you're having trouble with actually writing the case up, I recommend an approach that I call "Structured Argumentation" that I'll write about soon (and link back to here).

Trying out Elgg

2006-02-24T22:46:35Z

Lee Iverson :: Blog is my new blog on http://elgg.net, an ePortfolio system. I'm using it for my CSCW course EECE 519. I chose it because of the combination of blogging, tagging, selective sharing, and community definition and maintenance. I also have a student studying information sharing behaviours in a community of Elgg users. Let you know how I like it.

Tagging and Attribution

2006-02-06T19:32:25Z

A simple, but important point about tagging. If a social tag is unattributed, then it is next to useless. Think of "I tag Obj with tag" as an assertion that Obj "means" tag to me (whatever tag means to me). If this is done in a social environment, then without attribution and some ability to assess the meaning or trustworthiness of my tags (from your own perspective), the information that Obj has been tagged is problematic.]]> TagCamp, I gave a presentation on "Why tagging works" in which I suggested some reasons for tagging's success where previous attempts at social information organization have largely failed. Now, someone added a tag facility to the Wiki for the meeting, a good idea. But it was unattributed, and showed up in the sidebar when you visited a page. Sometime shortly after the meeting someone tagged my personal page with the word "blabbermouth".

Now, I don't mind a negative tag, but this was anonymous and thus kind of cowardly. And as far as I could tell it was inaccurate. A "blabbermouth" is someone who talks out of turn or reveals information they shouldn't. I don't do that. I may be a "bore" or a "loudmouth" (I happen to think best with my mouth open), but never a "blabbermouth".

But I also found cheated out of the opportunity to engage in a conversation with the tagger. What were they bothered by? Did I offend you or prevent you from speaking? I'm left hanging (and branded) without any possibility of correcting an unintended offence...

Now, not all anonymous tagging is of this character, but it certainly does leave me hanging.

Del.icio.us vs Google

2006-02-06T19:06:50Z

In A social analysis of tagging, Rashmi Sinha makes a few interesting points about the transition from solitary to social with the tagging experience. I tend to agree with this, but I would like to point out a few things I think deserve more consideration.]]> One there are a hundred or thousand responses to a search engine query the primary distinguishing feature of one search engine from another (besides speed) is the ordering of the responses. The main reason Google worked so much better than their competitors initially was the PageRank algorithm. What Sergey Brin and Larry Page figured out was that how many people are talking about you and what they are saying is more important than what you say about yourself. In essence, all the search engines that preceded Google did textual and structural analysis of the web page itself and created match values with key words on the page in order to calculate the strength of match with each query keyword. Brin and Page figured out that a link is primarily used as an attestation of authority and that the description of a page near its link location is more precise and significant than the page content itself. In other words it uses the "wisdom of crowds" to select pages that best match key words.

Tagging (in the del.icio.us style) has exactly this property. The primary differences are 1) that tags are open and attributed, and 2) that tagging is easier and more dynamic than building a web page. In essence tag systems mirror the pagerank structure of Google's system, but make the internal structures browsable and viewable directly. The consequence of this, and the primary strength of tagging systems, is that tagged structures are more trustable and harder to spoof or spam than search engines.

That said, they need to get better. I need to be able to express who I trust and especially who I don't. I need to express what information I trust from these sources and again what I don't (I may completely agree with a person's perspective on technology but abhor their politics). There's so much more that can and should be done with these systems...

Technorati Tags: del.icio.us, tagging, google.

del.icio.us and Recommended Tags

2005-11-02T22:55:04Z

On the weekend at TagCamp, I made a few points about the possibilities of del.icio.us and Flock for bootstrapping the development of rich, personal ontological models. In that framework, I made some unfounded criticisms of current del.icio.us practice that I should correct publically.]]> One of my main hopes for this stuff is that we are on our way towards allowing individuals to create contextualized information structures and then create systems that can mine these in social environments for community formation, grounded community ontological modelling, etc. In this context, as I see it, it is fundamentally important to allow and encourage individualized and potentially idiosyncratic tagging (or categorization) behaviour.

In that light, I was criticizing del.icio.us' "Recommended Tags" feature for its seeming push towards harmonization or homogenization of tagging behaviour. I was concerned that "recommending" tags based on how others had tagged an item would actively encourage decontextualization of the tagging activity.

Well, I was wrong. The "Recommended Tags" are derived from an intersection of tags others have used and your own tagset. In other words, you are encouraged to reuse your own tags when those have already been used to identify the resource by others. This seems to be much more about simply recognizing a natural concordance of your tagging choices with those of others. In other words, exactly the kind of "mining" of the social environment that I advocate. Mea culpa. Sorry for badmouthing you inappropriately Josh.

Technorati Tags: folksonomy, tagcamp, del.icio.us, tagging.

Yahoo 360 and Lucasfilms' Habitat

2005-04-05T22:15:14Z

Reading Yahoo! 360 through "The Lessons of Lucasfilm's Habitat". A great little article with another great little article in a link. Well worth the read.

Intermediation, REST & Bookmarking

2005-04-01T18:54:11Z

In The architecture of intermediation, Jon Udell suggests that there might not have been the need to "clone" del.icio.us (see de.lirio.us) in order to be able to experiment with the structure and services. He suggests instead the with the proper "intermediation" in terms of proxies etc. there should be ways to do things with del.icio.us that the designers haven't anticipated.]]> In a certain sense, I agree. The RESTful architecture of del.icio.us should make it really easy to do some interesting things with smart proxies and other kinds of services. In essence, the real hope an promise of the RESTful web services model (over SOA-like models) is the potential ease of aggregating services and really doing web services "right" with little cost.

The danger with this approach is the dissolution of services into a morass of intermediated systems. In essence, if the really useful information and services are all managed by these intermediated services then the benefits of "mining" the information are lost. In essence, a big part of the benefit of what I'm calling "object-out interaction" is potentially lost if the object reference information isn't somehow centralized. By that I mean, the opportunity to discover like-minded people and communities from the objects out to the people. I'm writing a paper on this at the moment, so more on this in the future.

Now, the "somehow" up there was also deliberate. Another nice thing about RESTful systems is the ways in which notification and update can be managed. It should be possible to have these intermediating services notify the central system in various ways to maintain the advantages of centralization. It is not clear how easy that is to do at the moment.

Finally, I do see a significant problem in this whole approach. Because of the need to mix representation (read data models) with presentation (read user-interface) in these RESTful services, it can be really quite difficult to provide a completely different UI as a front-end via such intermediated web services. Ideally, the real strengths of del.icio.us, which are in its data models, would be more or less directly accessible in a way that would allow for wide experimentation in user interface and service aggregation, without comprising the integrity of the original service. Hmmm... Definitely more on this later.

What is a Facet?

2005-03-18T19:38:52Z

Nils Soren Peterson in Semantic Tagging, an extension to a Group's Thesaurus asks "what is a facet?" in the context of an experiment being by Haiko Hebig with his del.icio.us tags.]]> He speculates that facets are like groups, but misses the point a bit. Facets are essentially like typed links, not groups. The best way to understand it is to think about the relation that a tag represents. Most tags have the meaning "resource X is about Y", where X is the tagged item and Y is the tag. In that sense we have an implicit "about:" or "subject:" facet. But we could have "kind:" facets representing the class of object or even facets tracking other metadata for an item "date-published:" or "site:".

In its simplest form a facet thus an aspect of the resource in question. For more, it is worth checking out the IAWiki: FacetedClassification.

Controlled Vocabularies Cut Off the Long Tail

2005-03-17T19:12:51Z

Bokardo � Controlled Vocabularies Cut Off the Long Tail makes a couple of important observations relating folksonomy to controlled vocabularies (e.g. domain ontologies). In particular he contrasts:

Discovery vs. Finding, or what I would term browsing vs. search.
Tracking vs. Prediction, although he doesn't quite abstract it in those terms.

In general, I agree with much of what he's pointed out, but I think he misses a few important aspects of the problem or generalizations of these observations.]]> In particular, there's a big issue around contextualized organization and decontextualized (or broadly contextualized) organization, and it's largely an issue of goals. Folksonomy approaches start from the point-of-view that you give people simple tools for organizing the information that they use or have used and this may or may not be useful for others. The controlled vocabulary approach assumes that a particular community has a common way of organizing the world and the vocabulary reflects that common scheme. Folksonomies attempt to support categorization in an individual's own context while controlled vocabularies assume a community context and assume that will be useful for all of its members (and possibly non-members).

The failures of the controlled vocabulary approach largely lie, in my mind, in this decontextualization of knowledge. I often look back on Jonathan Grudin's "Work vs. Benefit" analysis in assessing whether or not community technologies have a chance of actually working, and in the case of categorization, it is often more work for an individual to conform to a community vocabulary than to simply build and use their own, especially if the tools for doing so are extremely simple (e.g. del.icio.us or Flickr). Moreover the benefit (to the individual) accrued from using the controlled vocabulary is minimal. This is one of the two reasons that tagging is succeeding.

The other is simply the collapse of the horribly poor metaphor for organizing digital information that is the "Folder". And in this respect, it is not that it is an incapable metaphor or hard to grasp, just that it doesn't scale. It doesn't scale to a large number of items, and it doesn't scale to mutiple users. And unfortunately, I believe that the pure tagging approach to folksonomy development will fail for the same basic reason -- scaling.

Do Tags Work?

2005-03-07T22:02:22Z

In ongoing � Do Tags Work?, Tim Bray asks whether tagging systems such as Technorati or del.icio.us work? He suggested a simple piece of research that might find out...]]> I think there's something a little more going on here though. It's not "do tags work?" it's more of a "how far can they go?" In a simple sense it seems that keywords (i.e. author supplied tags) vs. tagging (i.e. community supplied tags) is really something that Google already proved. What the community thinks of you is just a better way to assess how relevant or important you're information is than what you think of yourself. And if you don't think that's what's behind Google's PageRank then go back and read the original papers again.

The real question (and this should be the questin with every proposed information management tool) is "how does it scale?" What happens when I go from 100 tags to 10,000? What happens when we have synonymy? How about conflicts? The "miserable failure" manipulation even points out how something as global as Google's approach can be exploited.

My answer? I suspect that the way out of this is a combination of organization of tags (probably semi-automatic rather than fully automatic) and some real models of testimony (i.e. I believe X to be true because Y asserts that X is true). In other words information management paired with social networking.

How could this play out then? First organiztion. Well, frankly I believe that RDF and ontological modelling are the right foundation for building and describing conceptual organization (and that is what tags are really about, named concepts). The real question is how to make the formation of these conceptual models accessible to real users and useful for finding and organizing information. I say that we should start from the bottom up: let me model my own knowledge and information resources efficiently and practically (i.e. find something to replace folders, desktops and simple taxonomies) and then find ways to coordinate these organizational systems with collaborators and friends and ultimately the communities we participate in.

The next, and possibly more important thing though, is to take testimony seriously. For example, I might want to know what the most popular sites tagged with "baseball statistics" or the most recent. Strangely, del.icio.us actually screws this up, since what I'm really asking for is a different ordering and not a filter. But, frankly, I'm much more interested in the opinions of a few people who I trust to suggest the best sources of information (e.g. Bill James). So, what I really want is both... I want to tell del.icio.us that "these are the people whose opinions (and bookmarks) I trust" and therefore give their tags greater weight in presenting things to me. I may want to be slightly transitive on this too, giving some, but less, weight to those they trust. And of course the opposite is useful too (e.g. "I don't trust doofus' tags") Del.icio.us does use this a teeny touch with the "ignore" feature, but that's a really broad broom and it prevents you from seeing stuff instead of merely downgrading it.

Now of course, testimony is contextual too. Frankly, it is very likely that I will agree with someone's technical recommendations and have a big problem with their politics, so that should be a consideration too... It is easy to see how far this could go and not really too difficult to do either. Anyone game?

Seb's Social Software Challenge

2005-03-03T07:18:30Z

Seb's Open Research describes a "Programming Challenge" to count links in deli.cio.us topics. There's some hope there, but I think we need to go in a different direction.]]> The problem with simple tagging systems like deli.cio.us and FlickR are that they just don't scale well. In essence, it's easy to organize things if you don't have that much to organize. It's when you get a lot of stuff that life gets tough. Think folders when you have 50,000 emails in your archive or a 500GB hard drive. What worked a while ago doesn't any more.

I'd suggest two things:

The addition of "testimony", a social component, to the tagging. In essence, there are people whose tags I trust and those I don't. Moreover, those I trust I probably only trust on some topics and not on others (e.g. I can agree with someone's software development practices without agreeing with their politics or religion).
Some way of organizing tags. It does seem bizarre to use the term "folksonomy" without having any of the characteristics of "taxonomy" (i.e. sub-categories). The addition of very lightweight tagging with a model of the tags (and some ways of making it consensual) would just kill.

Hunter S Thompson and Qualitative Research

2005-02-26T16:10:15Z

ongoing � HST is a good start on some Hunter S. Thompson reflections. I must admit, I have read little of his stuff and my clearest image of him is Doonesbury's Uncle Duke, but a few things that I heard people saying about him ran true, and I made an observation in my CSCW class on Tuesday that I'd like to share.]]> Someone being interviewed on CBC radio suggested that HST's greatest contribution to journalism was his very personal stance. He took the point-of-view that the things he was writing about were too important to try to adopt a dispassionate, "just-the-facts ma'am" reportage (i.e. a traditional journalistic style). Instead he loudly reacted to events, told everyone who could read what his own opinions and biases were, and told people what he thought about what his subjects. In essence, he personalized his writing in order to try to create an opportunity to personalize their own reactions and thoughts to his subjects.

So, how is this related to qualitative research? Well, traditional scientific epistemology is about adopting a depersonalized stance, a dependence on replicable evidence and theory and models based only on such evidence. The kind of knowledge being developed is decontextualized and generalized and dispassionate (think traditional journalism). Qualitative research takes the general stance that we also need contextualized and personalized knowledge (especially when we are studying people or the relationships between people and the world) and we need methods and standards for ensuring that such knowledge is useful. One of the foundations of qualitative methods is exposing the researcher biases, the context in which the evidence was gathered, and the methods used to achieve whatever generalization was possible. The key is that a reader of a qualitative study, rather than just trusting the results on their face and using them as needed, has enough information about the context used to gather the evidence and make the claims that he/she can assess for themselves. I'd suggest that these are exactly the characteristics that made HST's approach to reportage so different and engaging.

So what lessons to be learned? Does on need to get drunk to do qualitative research? Is a huge, all-encompassing sense of self and mission necessary? No, just a relentless, self-critical honesty with yourself and your readers. A complete openness in terms of the background and suppositions in your approach. And finally, a sense that it is the reader of a qualitative study who has to be able to decide for themselves whether or not your work is relevant or useful to them. Anything that you do in your approach to the work or description of it that inhibits this ability is reducing its potential impact and usefulness.

May Hunter rest peacefully now that he's been delivered from his demons and sense of responsibility.