o brave new world: Tagging

Tagspace Beta 2 Predux: On Some Pre-release Discussion

Sat, 31 Mar 2007 00:48:50 GMT

I was delighted (and I admit a bit concerned) to see that the cat's begun to get out of the bag and elicit commentary out in the ol' Blogosphere, with a pre-release post by our own Mr. Seven attracting the attention and comment of one Mr. McLaws.

Robert McLaws asked several questions:

What if I want someone else providing the structure? Will the owner of the blog or forum be able to define their own default view of their content?
What about popularity services? Tagging is great, and so are views based on those tags... but how do you know what other people are reading?
Are there going to be APIs tied into these services, or will it be a Microsoft-only thing?
Will this be a software product, a la Community Server, or a Microsoft-only thing?

To which I responded in the comments:

Popularity: we will ultimately provide many different filters and sorting mechanisms; popularity will, in our view, derive from the number of tags that are applied to a given resources. Do you have some other measure in mind?
APIs: Definitely, on the way. More details to come, but REST will be a part of it.
Software product: Not in our current vision of things, and not explicitly out of the question some time in the future; mainly we're about providing adoptable/adaptable services that others can use to build out their applications and customer experiences.

(I don't have a response on the first question, but perhaps my colleague, Brian, can address that--I'll ping him.)

Then Robert responded on the popularity question:

For popularity, I'm thinking something along the lines of Digg or Dell's IdeaStorm. Being able to vote for items that people feel are good, valid, or worth reading.

A: While I'd say we're less interested in producing a popularity contest along the lines of Digg (as I understand it), we are interested in being able to offer filters and sorting mechanisms that allow you to see what's received the most tags, what the most used tags are, and who are the most prolific "tagspacers." In our view, the very act of bookmarking something/applying a tag to it is an indication of interest--a vote of value, if you will. If it turns out that we should add more explicit "thumbs-up/down/??" rating functions because the people wish it, then we may very well do that. This is ultimately meant to be a service that you should have a say in developing.

There have also been a few other comments and questions on his post from readers I'd like to address here, this one from Massif:

How does this relate to the platform being used by Channel9 and on10?
It certainly appears to have many of the same features.

A: We're hoping that, if successful, specialized community sites like Channel9 and on10 will see the value in what we're offering, and adopt our services to become part of a much broader Microsoft community of sites. Last I heard from my manager Bob was that the folks responsible for those sites have shown some interest... we'll see. But to be clear, what we're planning on offering is all about distributed and federated community services that eventually ANYONE out on the Web could theoretically leverage to serve their own needs and particpate in a broader community. The way this could work with Tagspace, for example, is that if Site "Z" (or whatever) wants to tap into our "cloud" of members, tags, and tagged resources and contribute their own, they could easily do so by registering as a namespace in our system. They could then implement custom or default controls that would display just those items/tags/people within their domain and/or any related domains and/or the global "tagspace." Further, tagging done in their domain would also be aggregated and available in the "cloud." There would be very flexible scoping options for constituent namespaces.

And this one from Corinne:

From the bit of experimenting I have done, I would think that maintaining the tags appears as though it would be a spam headache. For example, when selecting a URL, i.e., http://microsoft.com, it then becomes a major exercise to edit the tags to obtain anything of value. The current tag list for Microsoft Corporation:
Tags: __WowIcanTypeAbsolutelyAnything, <script>bbb</script>, AJAX, asdf, asdfd, Beta Office 2007 issue, bill gate, Blogs, c, C++, CanITagThis?, developer, downloads ie7, entertainment, Excel, Flickr, food, fun, Games, gateway, general information, groove, headlines, home, home page, homepage, IM, internet, Internet Explorer, ISA server, microsoft, Microsoft Egoism, Microsoft Home Page, mobile phone, mobile phones, movies, MS, MS home, MSCOM, music, pics, portal, products, profanity, rachel, radio, sam garmon, software, SQL, Sql server, T, talorp@microsoft.com, telephone, test, Travel, TV, videos, visual studio professional, wireless, wow, xp
Misspellings such as "Outllok" will also contribute to uselessness.
How do you prevent someone from embedding dangerous code in the tags?

A: Corinne is right to point out the potential (realized in the current Tagspace) for, um, "tagspam," to perhaps coin a word. This points out one of the quintessential paradoxes of tagging: the little-'d'-democratic virtue of the technology--whereby anyone can add value (or not) as they see fit by the choice of their labels for things that interest them--is prone to spoil itself through its very openness and lack of structure. The perhaps justifiable fear is that a system that's intended to cut down on the noise of the Web just contributes to it, the chaos of an open system thus ruining it for everyone. (Hmmm... sounds a bit like the Web herself.)

Now, there are ways to cut down on the noise, but you have to be careful to strike a balance. For example--and we haven't yet implemented this--you can simply limit the number of tags to display on any given item to those that are most popular, all others falling off on the "long tail." Over time, then, once you have a critical mass of participation and tags, the noise gets shunted to the bottom of the stack, simply because very few people would care to muck up the pool with seemingly meangingless, certainly bizarre, and arguably useless entries like "AllYourBase" or "yup." But by doing so, the risk is that you end up reinforcing terms and patterns of thought by focusing on what's ostensibly most "popular." Most used, to be sure, doesn't necessarily translate into most useful.

So it is clearly a question, and we're very open to your suggestions on how to answer it: how do you, with appropriate sensitivity and restraint, allow for the fullness of people's expression, enabling the discovery of all points along the "long tail," while at the same time ensure that the experience doesn't become "a tale / Told by an idiot, full of sound and fury, / Signifying nothing"? Ay, there's the rub (pace, Messrs Macbeth & Hamlet).

Now as far as embedding scripts are concerned, we're taking steps with the refresh to ensure that such things will be stripped out or disabled, principally by putting limitations on the number of characters in individual tags, and the number of tags to be applied to any one item. Let us know when the time comes how this works/doesn't work for you.

Ok, *whew!* That's all I've got in my brain pan for the moment. And my apologies to Robert for diverting the discussion over to my blog: I needed a bit more space to type-type-type out my thoughts and responses.

Thanks, All--

- dave

Using Tags to Structure Microsoft Community

Sat, 09 Sep 2006 02:28:58 GMT

Imagine if you will a new Web experience on MSCOM based on all this Web 2.0 craziness. Well that's exactly what we're trying to put together over the next several moons. My mission is to help define new tagging and collections experiences for Microsoft communities, starting with select properties on Microsoft.com.

One way to try to think about this is to imagine URL structures built off a hypothetical community.microsoft.com domain (don't try going there... it doesn't exist yet!). Here's what I've come up with, in perhaps belabored detail--but do indulge me some, I'm trying to think this through (and forgive me, as well, for the del.icio.us-ness of it all).

Here are some of the lists we’d like to enable for an individual:

http://community.microsoft.com/davemorehouse = my tabbed “portfolio” page where members can see (if made fully public) my activities, tagged items, collections, people, groups I belong to, etc.
http://community.microsoft.com/davemorehouse/items = the default view of my list of tagged items, collections, people, groups, and tags
http://community.microsoft.com/davemorehouse/items/tags/foo_bar = the list of my tagged items, collections, people, and groups corresponding to the tag “foo_bar”
http://community.microsoft.com/davemorehouse/items/collections = the list of all the collections I’ve created (comprised of items, people, other collections, and groups)
http://community.microsoft.com/davemorehouse/items/collections/tags/foo_bar = the list of all collections I’ve created corresponding to the tag “foo_bar”
http://community.microsoft.com/davemorehouse/items/people = the list of all the people I’ve tagged
http://community.microsoft.com/davemorehouse/items/people/tags/foo_bar = the list of all the people I’ve tagged who correspond to the tag “foo_bar”
http://community.microsoft.com/davemorehouse/items/groups = the list of all the groups I’ve tagged
http://community.microsoft.com/davemorehouse/items/groups/tags/foo_bar = the list of all the groups I’ve tagged that correspond to the tag “foo_bar”
http://community.microsoft.com/davemorehouse/people = the list of all the people I’ve chosen to “follow” or “stalk”
http://community.microsoft.com/davemorehouse/groups = the list of all the groups I’ve chosen to create or join

We also need to think about how tags should be represented as applied in specific contexts, like in blogs, forums, TechNet, or other sites. Initial thoughts on how this could be structured:

http://community.microsoft.com/items/tags/foo_bar/site?=forums.microsoft.com = the list of all the items tagged “foo_bar” in the site forums.microsoft.com
http://community.microsoft.com/items/tags/foo_bar/community?=MSdevelopers = the list of all the items tagged “foo_bar” in the MS Developers community (comprised of multiple sites)
http://community.microsoft.com/davemorehouse/items/tags/foo_bar/site?=forums.microsoft.com = the list of all items I’ve tagged “foo_bar on the forums.microsoft.com site

Here’s how RSS feeds and OPML might be done:

http://community.microsoft.com/rss/davemorehouse = the feed that corresponds to all my activities I’ve elected to make public
http://community.microsoft.com/rss/davemorehouse/items = the feed that corresponds to all of the items I’ve tagged
http://community.microsoft.com/rss/collections = the feed that corresponds to all the collections I’ve created/will create
http://community.microsoft.com/rss/collections/to_read = the feed that corresponds to the specific collection I’ve labeled “to_read”
http://community.microsoft.com/davemorehouse/items/collections/foo_bar.opml = the opml file for my collection named “foo_bar”
http://community.microsoft.com/davemorehouse/items.opml = the opml file for all my tagged items
http://community.microsoft.com/davemorehouse/items/collections.opml = the opml file for all the collections I’ve created
http://community.microsoft.com/davemorehouse/items/groups.opml = the opml file for all the groups I’ve tagged
http://community.microsoft.com/davemorehouse/items/people.opml = the opml file for all the people I’ve tagged
http://community.microsoft.com/davemorehouse/people.opml = the opml file for all the people I’ve chosen to “follow” or “stalk.

Or, perhaps the structure should continue to be something like: http://community.microsoft.com/opml/davemorehouse/people, etc.

Then there are the global pages, not specific to users:

http://community.microsoft.com/items = the default view of all the items, collections, people, and groups that members have tagged, sorted by date saved, most recent first
http://community.microsoft.com/items/collections = the list of all collections that members have created
http://community.microsoft.com/items/tags/foo_bar = the list of all the items corresponding to tag “foo_bar”

etc.

Finally, there are the tag and user combinations that should be possible:

http://community.microsoft.com/davemorehouse/items/tags/foo_bar+bar_foo = All the items I’ve tagged corresponding to the tags “foo_bar” and “bar_foo”
http://community.microsoft.com/davemorehouse+bobreb/items = All the items that both bobreb and I have tagged
http://community.microsoft.com/davemorehouse+bobreb/items/tags/foo_bar+bar_foo = All the items that both bobreb and I have tagged with both “foo_bar” and “bar_foo”

... and so forth.

Any thoughts on that? Any obvious logical gotchas?

Playing Nice, or Opening a Can of Worms?

Mon, 21 Aug 2006 20:30:44 GMT

Korby and I were just now discussing the marketing plan he's working on for the projects we have planned for the next few fiscal years. He indicated to me that he was getting into the sections related to threats and competition, which got us to talking about the nature of competition in the social networking/computing space. Without going into any detail about who he saw as our potential competitors, I voiced the point of view that in order to be successful with what we have in mind for social bookmarking and/or tagging (the latter being a component of the former, the two not being mutually exclusive, though often confused), it might not be a bad idea for us to make it easy for users of our system to bookmark or tag stuff on Microsoft-related sites, and at the same time, allow them to send their tagged/bookmarked pages to their favorite social bookmarking service (e.g., del.icio.us, furl, or myriad others).

The basic problem as I see it is this: those hip to social bookmarking and/or tagging--those whom we'd want to attract first and foremost with any tagging solution we might provide--likely would be turned off by a system that didn't take into account the fact that they likely already had their preferred bookmarking solutions. In other words, if we offered another stand-alone tagging and/or social bookmarking service just for Microsoft-specific properties and third-party communities, we'd just get lost in the crowd and wouldn't be perceived to have much independent value. I mean, why bother participating in another bookmarking and/or tagging solution if it didn't tie into what you already used? The user benefits would have to be extremely compelling to say the least.

So what if we offer a level of integration with popular bookmarking services already out there, assuming this poses no significant legal problems? What if, say, you could tag and/or bookmark MSFT properties and related communities--derive whatever attendant benefits from the overall solution we'd offer--and at the same time be able to maintain your del.icio.us links and tags? Is that a good idea? Should we go the route of Diigo and others that permit posting to multiple bookmarking services? Maybe the answers to these questions are obvious, and, like, no duh, of course we should play nicely with others in social bookmarking and networking communities... After all, if Live Writer can integrate with various blogging platforms, why can't we integrate with various third-party social bookmarking platforms?

You tell me--what do you think? What are the pitfalls we should watch out for? What would be a shortlist of bookmarking/tagging services we should try to align with?

______

8/23 addendum to this post. Here's what Diigo offers, for reference:

LibraryThing, Tagging, and Forums

Fri, 11 Aug 2006 21:06:05 GMT

LibraryThing has implemented a means to tag posts inline, in their new discussions solution called "Talk." (Thanks to Joho the Blog for pointing this new feature out.) Basically, the idea is that as you write up a forums post, whenever you make reference to a book or author, you simply enclose the book name in single brackets--e.g., [The Assassins' Gate: America in Iraq]--and the author name in double brackets--e.g., [[George Packer]]. This signals to the system to do a lookup on the bracketed text and make the required match in their database and automatically insert the hyperlink within the post to the book or author that's been tagged. Further, since posts are tagged in this fashion, it improves the discoverability of and navigability to content that's potentially buried in long--or even old--discussion threads.

Now, Microsoft has technical forums. Consider Tim Spalding's argument that "forums are broken" because, in his words:

There's too much to wade through, and most of it isn't really want you want. Creating subforums for "Romance" or "Current reading" helps, but beg the question of appropriate organization in a fixed, confining way.* Divide the word differently? You're out of luck! And if a community forms around the preset topic, getting to know the other members of the community is a long, and not necessarily pleasant process. Because they require so much time and energy, traditional forums tend to favor "loudmouths" and worse. And the whole enterprise spoils faster than milk. Nobody digs through a year-old "Mystery" forum looking for posts about a midlist author who could equally well fall under another genre. Most don't allow you to reply to old posts. What's the point, when nobody else will end up reading it?

Let me just raise this question: do Microsoft forums suffer from these kinds of issues? Arguably, yes, to one degree or another, and I think that relying on a "silo'd" search solution to discover threads and posts of interest that might be buried in long and potentially old discussions is severely limiting.

Perhaps what we need to do is consider a tagging solution along the lines of what LibraryThing has done. We could potentially use LibraryThing's system of bracketing author keywords, but that wouldn't necessarily be the most intuitive thing for forums users to do to flag their tags (though conceivable, if key influencers got the hang of it, and others decided to follow their lead).

I imagine it would work something like this:

Take this recent post from the Architecture General forum. Using LibraryThing's methodology, an author could mark up their post for keywords like this:

In the [admin module] of the application, we have a provision to [add a user]. On adding a user through the application, the user is added in 3 places, namely in [Active Directory (AD)], [SQL Server] [database] and in [Sharepoint] database. All this is a part of [ASP.NET] code with no transaction happening through out the user creation.

Now is there any way to impose a [transaction] in this process? That is, until all three AD, Sharepoint and SQL Server databases are up and running and the creation of user happens with no errors, the new user must not be committed in all three databases.

This is very critical as it causes inconsistency if the user is saved in one database and an error occurs in another.

Or, possibly more user friendly, would be a simple tagging highlighter that would be a part of the WYSIWYG controls in the post interface. Another tack we might mull over is a simple button ("Tag This Post") in the control set of the post interface that would pull up a window that would automatically suggest a list of keywords drawn from a text-miner-type application--a list that the user could subtract from or add to as he/she desires. Some of the keywords could also be drawn from context, e.g, since the post cited above is located in the Architecture forum group, it would automatically receive "architecture" as a suggested tag, along with "msdn forum," or other contextually relevant markers. Perhaps these latter would be automatic system tags, uneditable by the user.

In any event, what LibraryThing is doing with inline tagging is very interesting, because it begins to show how threaded discussions and comments (like what you have attached to blogs) can be user-attributed, with minimal development and pretty straightforward user requirements. Once systems like this are in place and, of course, actively used, we at Microsoft.com can do a much better job of making user-generated content in contexts such as forums and blogs much more discoverable and so extend its value to the community.

What are other tagging solutions for threaded discussions and comments that we might consider?

What are some potential applications of tagging?

Fri, 28 Jul 2006 21:37:29 GMT

Been noodling on what value users could derive from a tagging solution for Microsoft Web properties, and it seems to me that abundant possibilities present themselves. I've come up with four big buckets:

Facilitating the discovery of information
Facilitating the discovery of like-minded (or even differently-minded) people
Facilitating the discovery of people's participation
Facilitating collecting and sharing information

Now, what are some examples of each? Here's a good selection of scenarios we should consider:

Information Discovery

Show me all resources tagged with this tag
Show me all tags related to this resource
Show me all tags related to this tag
Show me all resources related to this resource
Show me the most popular resources
Show me the most popular tags
Show me the most recently used tags
Show me the most recently tagged resources
Show me the resources tagged with both this tag and that tag (junction)
Show me the resources tagged with either this tag or that tag (union)
Show me all the tags used by this person
Show me all the resources tagged by this person
Show me all the resources tagged by either this person or that person (union)
Show me all the resources tagged by both this person and that person (junction)
Show me all the tags used by both this person and that person (junction)
Show me all the tags used by either this person or that person (union)

People Discovery

Show me all the people who have used this tag
Show me all the people who have tagged this resource
Show me all the people related to this person based on tag affinities

Participation Recognition

Show me all the most popular people (i.e., ones who are "stalked" or subscribed to)
Show me all the people who have recently tagged
Show me all the people who tag the most

Information Collection & Sharing

Let me see all the things I've tagged and all the tags I've used (private)
Let me show the things I've tagged and tags I've used to others (public)
Let me subscribe to and share feeds for all of the above

Clearly these use cases will have to be prioritized to some extent based on technical limitations. As I've come to understand, querying across large data sets to provide the Boolean joins between person, tag, and/or resource--e.g., show me only the people who have used this combination of tags--is a complex and costly matter. And in any event, those kinds of scenarios are probably much rarer and much less useful to people than the basic ones that allow them, for instance, to see all the resources tagged by this particular person, to say nothing of simply being able to keep track of their own resources and tags.

Are there others we should be considering in building a tagging solution for Microsoft Web properties? What would you prioritize as the most important scenarios for retrieving information via tags?