connecting*the*dots

Thanks, Chris & Tara.

If you happen to be someone who thinks all this 2.0 hub-bub about social tagging and meta-data is confusing, I’ve found the perfect domain for us to reverse engineer together.

In order to get us on the same page, why don’t you first hop on over to Musicovery — an online radio site with an extremely interesting interface — and play around for a bit (be sure to explore all of the feature found on the controller).

Just don’t forget to come back! I promise that we’ll have some fun and you might even learn some geeky information architecture stuff.

Welcome back.

Okay, so how brilliant was that experience?

I don’t know about you, but discovering music based on my current mood fills a huge void in how I currently listen to music. Before discovering Musicovery, the closest I could come to replicating such a dynamic experience in iTunes was by creating a playlist for a specific genre and shuffling the playback.

And that just doesn’t do it for me. (more on the genesis of genres later)

Essentially, everything that Musicovery is doing is made possible by leveraging the relationships between meta-data applied to discrete information objects. So, are you up for digging further into the underpinnings of this puppy to figure out how it works and possibly come up with a few meta-data driven enhancements to the current user experience?

I’ll take your silence as a yes. Alright, let’s get to it then.

Old School, Structured Meta-Data

Deconstructing music (as an information object) is pretty straight-forward, as each song comes with standardized attributes that neatly fit into industry-wide delivery and marketing mechanisms (which were established well prior to the explosion of the dynamic nature of the web).

Okay, first, let’s list the most commonly exposed and explicit attributes of a song. My top six would be:

• Artist name
• Song Name
• Album name
• Release Date
• Track Length
• Genre

Now, while the first five attributes are all explicitly defined — the artist’s name is the artist’s name, etc. — the sixth attribute (genre) is only explicit when viewed through the lens of the music industry’s nomenclature levers (a song that I consider to be hip-hop, someone else might call rap, while the music industry itself might label it as pop).

By managing the evolution and edification of genre nomenclature, the music industry uses these silos to market acts with a much greater degree of certainty in matching the expectations of the customer because the music industry is creating those very expectations themselves through this process.

Deep, huh?

So back to deconstruction; let’s see how Musicovery is leveraging these primary attributes (if at all):

• Each song displays the artists name
• Album name isn’t exposed
• The controller interface allows the user to narrow results by decade or specific year based on the release date
• Track length isn’t exposed
• Genre is displayed prominently in the controller as the primary filter of returned songs

Two of the six most prominent song attributes aren’t being used, yet there’s a preponderance of controller functionality left to discuss.

Something else is going on.

Meta-Data In The Digital World

The aforementioned attributes of the song object have been around forever; they are the core identifiers for a song and always will be. As I mentioned before, the music industry has become extremely efficient in managing the relationships between these attributes across an expanding universe of songs — it’s their lifeblood. This particular set of meta-data fit the strategy of the analog age of information — where meta-data was constrained to the physical dimensions of the record’s liner notes or the pages of an industry magazine.

Now, in the Information Age, there are truly no limits to the amount or types of meta-data that can be generated; the only limitation — from a practical, business perspective — would be in how these new attributes fit into the domain’s value equation.

So, because the folks behind Musicovery have focused on creating a radio application that exposes music in particular ways (other than shuffled programming or human dj’ing), it’s a solid bet that they’ve expanded upon their meta-data set.

The Nitty-Gritty Attribute Model

In order to return a song by clicking on a specific spot in the mood or dance interfaces, the quadrants need to be explicitly defined to hook up with corresponding attributes applied to songs in the Musicovery universe. So what type of attributes would we need to add to each song? Here’s one approach:

Mood Interface

• Dark to Positive attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• Calm to Energetic attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)

Dance Interface

• Dance (-) to Dance (+) attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• Tempo (-) to Tempo (+) attribute scale (-5-,4,-3,-2,-1, 0, 1, 2, 3, 4, 5)

The range could be much more refined than 11 data points — theoretically, it could be as refined as equating to the number of pixels that reside in the actual interface — but due to the current size of the song universe (it seems limited, as I get repeat results somewhat often) and the already subjective nature of assigning such attributes to songs, this degree of differentiation would probably suffice.

Now, let’s take the mood interface and chop it up along these lines to visualize how each song could be found in this manner:

That’s pretty much it.

So while there are numerous choices one could make in the presentation (depending on the size of the song universe, the visualization would span out to neighboring squares to present a full return, etc.), in order for a song to be accessible by any aspect of the Musicovery interface, each song object would simply need to have the following structured data applied to it:

• Artist name
• Song Name
• Release Date
• Genre
• Dark to Positive attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• Calm to Energetic attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• Dance (-) to Dance (+) attribute scale (-5,-4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• Tempo (-) to Tempo (+) attribute scale (-5-,4,-3,-2,-1, 0, 1, 2, 3, 4, 5)
• A Billboard ranking (0,1) in order to display whether the song was a hit or not

Most of these data points could be data entry for a trained monkey, but the scaled meta-data is such a subjective determination that the resulting experience will vary from person to person.

Aside from scouring for top, authoritative talent like Kennedy (eh, for early 90’s music) and pay her thousands upon thousands of dollars to “moodize” and “dancize” each song and then splash her grill on the interface to pimp the brand, what else could we do to improve the resulting experience?

If you know me at all, you know where I’m going with this.

Why have only one person or team from one domain attributing mood or dance settings to all music, when the openness of the web has already proven models for empowering each user with the ability to add their own meta-data to the mix if they should chose to do so?

Open Up The Gates

Way back in the day, Launch.com (now Yahoo! Music) was the king of the internet radio scene. And while I dug being able to subscribe to other user’s services through their social network, my favorite feature, by far, was the ability to rate my music on a 0 (never play again) to 100 scale, in increments of 1.

Sure, maybe 101 levels was over the top, but future playback of my favorite music was amazingly accurate. Now, what if Musicovery allowed this same type of two-way interaction?

Here’s an example scenario:

I just clicked on the mood interface between the energetic and dark nomenclature. The first song that returned was Joe Cocker, With A Little Help From My Friends.

Really? Dark and energetic? I don’t think so. But as it is, I can’t affect the centralized intelligence of Musicovery. I just have to take their recommendations at face value.

Now, what if we were to add user input into the song interface?

Once we added our perspective on mood, the system could return the results to the information object and use the input in two ways.

• The meta-data could be lumped into all user feedback to present a more representative mood interface — the wisdom of the crowd if you will
• It could also be used to present personal mood results, from a toggle setting in the interface

If the song universe was large enough, we could add a similar rating control that Launch employed, so not only would our mood expectations be met, we’d hear our favorite songs more often as well.

Fun stuff.

Tonight @ 7pm in Congdon Hall, Room 138 at High Point University (Directions), John Ford and myself will be rapping about this little activity called blogging.

If you’ve heard about it before, but don’t know how blogging can assist you as a small business owner, an activist, a writer, etc., come on down and get both the back-story and the 411 on how to publish to the internet.

And if you’re already a blogger, well, come on down and live-blog our presentation!

(originally uploaded by wmchu)

Just as we were getting used to how folksonomies can help us find relational information, ‘dem darn kids take it to the next level.

Long gone are the days when protesting corporate bullshit was limited to groups of people gathering on the street outside of a main office. Nowadays, you can protest by simply dropping a single word into the workings of the retail experience itself.

Check out what DefectiveByDesign is doing:

How passé is crafting a product review now that you can group multiple sucky products that share a common sucky trait with a few key strokes? Why tag your frustrations on your blog, when you can hit the fuckers where it hurts the most — in the virtual aisles and checkout lines themselves?

Excuse me while I head over to Amazon to spread my love of hating DRM.

UPDATE: Tag-daddy, Thomas Vander Wal, makes a profound statement on my flickr comment thread.

(via BoingBoing)

Lisa Scheer and I spent a few hours over at M’Coul’s Pub yesterday, melding minds over how to best use the web to expose her amazing eye to a larger audience and start a conversation about her passion.

Enter tagging.

After a few hours of exchanging philosophical approaches and dissecting interfaces, Lisa left with laptop in tow to start exploring her new sandbox.

Her castle is going to be dope.

Can flickr be any more fun without spinning in circles before exploding into fiery, shimmering glitter dust?

For those of you not in the know, geo-tagging is when you apply specific (or general) geographical tags to an object in order to identify its location. flickr has done an amazing job out the gate with this puppy, as the drag and drop interface is so good, so very easy to use.

(click here for a full-sized interface screenshot)

I’ve spent this entire evening digging back through my photostream, eyeballing maps and looking up the addresses of specific places where I took my shots. Some are easy to find (my house, M’Coul’s), while others are a bit of a challenge (wedding pictures, scenic shots), but it’s a fun exercise either way.

My question to Stuart and crew: This is going to become socialized at some point, right? (UPDATE: The map just appeared in my Explore tab! More here.)

I mean, how fresh would it be to be working your map and easily flip from how you’ve experienced a location to how someone else has? Essentially, take the concept behind the tag globe icon and apply it as a metaphor within the map interface, opening it up as another exploration tool? (I realize that I’ve just described a lot of the functionality of Plazes, but it already relies on people uploading geo-specific flickr images of hot-spot locations to their interface… hm, another Yahoo! acquisition, possibly?)

The Business Of Mashups

When I interviewed/presented at A9 last June, they were in the midst of that highly publicized “send a college student around in a van to take pictures of every block of every city” campaign. The idea being that seamless visual context of a business location on a Yellow Page business interface could be both useful and fun.

Well, sure, but the most useful? I approached the interface challenge from a bit of a different angle.

My presentation ended up clashing with what I perceived to be their primary context scenario for the product (people finding particular businesses with city block pictures). I argued instead, focus first and foremost on improving Yellow Pages search results and try to get businesses to “tag” their particular inventories to expose their goods to the A9 engine. Simply put, lead with the most useful user scenario, not with the eye candy of street scenes, which can always come later.

Now, flickr is, and should be, all about enhancing eye candy (finding it, sharing it, etc.); enabling people to find geo-specific businesses that have what they need is someone else’s business model.

See where I’m going with all of this?

Imagine how sick of an API this geo-tagging feature would be for a Yellow Pages product — one completely optimized to the teeth with a killer business tagging interface, providing exponentially more degrees of findability than simply scraping language from the business name, description and reviews found on the business interface itself?

Say a kid, fresh on campus, is looking for a local Chinese food restaurant and stumbles across the smartly exposed collection of quarter-mile range of images on the business interface of a Yellow Pages service. I can imagine the following conversation busting out:

Dude, check this out! ‘Swallow Balls‘ Haha. I’m getting that for Joe, he’s such a ball swallower. Ha! Oh man… they even serve scorpion? Okay, we have no choice, grab your chopsticks, we’re so there!”

Viral goodness of flickr madness; good for you, me and Mr. Chen.

Gnar, dude.

In my long-term quest to become more localized with my business ventures, I recently partnered up with John Ford at Aldenta to design a web strategy for one of his clients: Louis’ Healthy Breads.

Louis is a local entrepreneur who pours his heart and soul into his amazing line of health food products. Prior to our initial meeting, John mentioned that he was passionate about his product, and after meeting him, I have to say that he’s as authentic as they come.

Okay, a gear clicked.

After speaking with his daughter, Ann (his soon-to-be-hired editor and marketing person) and hearing about his son, Grant (the Chief Chef of the operation), my original plan began to grow some serious legs.

Yep, we’re going to create a Louis’ Healthy Breads blog.

Logistically speaking, Louis’ hip to email, but he’s not technically savvy and he’s always on the go — visiting stores across the east coast and beyond, trying to land retail deals for his product. The best thing about his day to day is that he’s constantly meeting interesting people who are always sharing personal stories, ranging from pure testimonials to insight into their health issues and fitness goals. Hey, he’s a salesman from the heart; a great candidate for simple posting, but not necessarily the archival aspects of tagging or the community aspects of linking out.

Louis is already doing the conversational work of a blog, but his “interface” doesn’t have built-in permalinks, comments and trackbacks (heh). Our job is to figure out the best way to corral his conversational personality and guide it into a realm with a searchable past and a participatory interface.

As a next step, I sat down with Ann last week to discuss a blogging strategy for the company. We both realized that Louis’ strong suit is writing, so we’re going to play to his strengths and limit his role as much as possible, keeping his publishing responsibilities focused by sending emails to Ann for editing and posting. We applied the same thinking to Grant, who will be able to provide a “behind the scenes” look into the baking process, leads on new ingredients, etc.

By heading down this road, Ann has implicitly agreed to become the blogger, editor, curator extraordinaire. Essentially, she’ll be editing posts from Louis and Grant, applying structured tags to the posts, linking out to related conversations across the blogosphere, participating in related conversations, managing comment threads and tracking related information in her RSS aggregator.

The idea that I’m trying to impart is that we don’t need to “segment” people into groups or “target” them like they’re deer in order to “push” product. We already know that the majority of people who are drawn to Louis’ Healthy Breads share certain interests, desires; call them attributes if you will. Health enthusiasts gravitate toward the product, yet their reasons for doing so could be as diverse as a fear of heart disease and diabetes to wanting nutritious fuel just prior to a marathon.

Ann is going to make it her personal mission to explore these existing communities, enter the fray of their current conversations and build relationship with actual people (not segments). Once she finds these resources, she’s going to start tracking them for interesting conversational nuggets to point to and contextualize from the LHB blog. Then, and only then, will she (and LHB) come off as credible participants in the communities they wish to serve.

Checkout the above workflow (228k | .pdf) that I put together to help get them off the ground. Can you think of anything that I might have overlooked?

Dave Winer has expanded on his “It’s the users, dummy!” statement and I couldn’t agree with him more:

There’s actually a neater solution, especially if you’ve put a piece of software on the user’s desktop to facilitate uploading and editing of the data — keep a copy of all the data on the user’s desktop, and just mirror it in your web app. There goes the problem (or is it an excuse) that your competitor would be using your CPU cycles to grab a copy of the user’s data (with the user’s permission, I should add, you need a username and password to get access, so the argument that they’re protecting against scrapers and abusers doesn’t hold water).

Following the Beyond Broadcast conference in May, Nate Aune and I began jamming on a similar concept; something we loosely called myTag.

(click for entire .pdf)

The major difference in our approach is that we’re trying to create a “piece of software” (actually, an online service) that can work across all online services, serving as a meta-data hub for all personally tagged information objects — blog posts, photography, video, audio, social bookmarking and possibly service metadata, such as Amazon tagging.

Unlike Dave’s example, we would scrape external services for newly updated, tagged objects. The goal is to centralize people’s meta-data and provide ownership of said meta-data, not to interfere with people interacting with these decentralized services. The scraping would only occur when a person accesses myTag to review their current tag universe, so the impact on external server CPU cycles would be innocuous at best.

Dave’s idea focuses specifically on the data editing and management issues that exist when users attempt to move their data across existing services:

With a local copy, the user can point any service at the data, and it can suck up a copy, and the competitor’s app would run on the user’s desktop too, using their (abundant) CPU cycles. The vendor’s server (in this case Flickr) wouldn’t even know that a copy of the data has been made, and since it’s the user’s data, that’s exactly as it should be.

Yet another reason to use rich clients. I use Flickr Uploadr, always. It’s just a bit easier to work with than the browser-based method of uploading, and that bit of easiness has proven to be worth it. Then of course the competitor has to offer a desktop tool as well. We do it with the OPML Editor. The server components, the directory browser, blog renderer, work with a copy of the data, the originals reside on the user’s machine. It also protects against a system failure, or a company failure.

I completely agree with his ownership point regarding meta-data, and his perspective of safeguarding information objects from system or company failure is extremely valid as well.

So how could we extend his concept of locally managing data (both the information object itself and its meta-data) across same-type services (flickr, zooomr, riya, etc.) to include culling meta-data across different-type services that leverage tagging (del.icio.us, YouTube, flickr, WordPress, etc.)?

See, the reason we’re sketching a thin client is because our primary goal is to enable individuals to be able to review and curate various slices of their own concept terminology — meta-data or tags — as they’ve been applied to information objects over various services and periods of time.

The way I look at it, an aggregate of tags can serve as a looking glass into the personal linguistic structure of each of us, as we make explicit choices when applying specific concept terms to our objects. As competition to flickr and YouTube enters the market, our POV’s will undoubtedly become further dispersed across the web, increasing the findability of our objects, yet conversely affecting our own understanding of our perceived output.

If we’re going to arm citizens with media tools, then we need to provide intuitive, smart representational interfaces for accessing and modeling our own strategic output. Why? Well, we need to be on-point, constantly iterating our understanding of our own perspectives and biases as we venture further into producing our own media.

Otherwise we fall into the same trappings of the mainstream media.

An example… take the limited nature of my tag cloud on this blog as an example. Click on a term, such as Greensboro, and a narrative will unfold over the period of time that you choose to explore and read. While it’s useful to understand more about my relationship with and perspective on Greensboro, the cloud doesn’t include my photos tagged with Greensboro, nor my video clips tagged with Greensboro.

Citizen media operatives need an centralized interface to access decentralized information objects. From my perspective, the value of these interfaces is huge — both to the content creators and potentially to the content consumers.

The first two scenarios I mapped out in the above sketch were for searching and browsing ones existing tag library. Any other primary scenarios jump out at you?

Bill Readings introduced me to linguistics back in my undergraduate days at Syracuse University. It was a low-level Critical Theory class, not enough knowledge to rest a proper degree upon, but that wasn’t Bills concern. He just wanted us to listen and think.

Bill had a wonderful way of illustrating his teachings — placing our 19 year-old minds into comfortable arenas where we could casually move towards comprehension, eventually grasping the core concepts of deconstructionalism and linguistics he tossed about with ease.

After choosing Blade Runner as an explicit assignment for visual deconstruction, and his daily, illustrative call-outs of us numskulls to apply a “bit more apperception to your day-to-day existence,” I’d have to say the strongest, most visceral lesson that stuck with me was his conversation around the English word “tree” and the Spanish word “arbol.”

An Attempt To Share Knowledge

To monolingual, English speaking folk first exposed to the authority of the Spanish translation, the inherent belief is that the two terms (English and Spanish) are perfect representations of the signifier, “tree”… which is wrong.

The signifier of “tree” is more akin to your personal mental model of the physical representation of:

original photos by icathing and Melete

Viewed through the lens of semiology and linguistics, we cannot absolutely assert that tree = arbol, because the signifier of “tree” has a unique representative interface to each of us, as does the percept of the translation of “arbol.”

Our individuality is too explicit to absolutely relate to explicit terminology.

Or put into political terms, in this society of modern constructs — one that consistently nudges us towards silos of absolute knowledge, relationships and definition — we are presupposed to assign relative constructs of our world to get by, based on what, in essence, is an aggregate misunderstanding of our own individual cognitive processing.

Back to the tree example; Roland Barthes on Saussure:

Until he found the words signifier and signified, however, sign remained ambiguous, for it tended to become identified with the signifier only, which Saussure wanted at all costs to avoid; after having hesitated between some and seme, form and idea, image and concept, Saussure settled upon signifier and signified, the union of which forms the sign.

Nowadays, whenever I stumble upon a conversation about knowledge and structure — such as Are trees natural? over at David Weinberger’s blog — the information architect within me rests in a state of nirvana, coaxed into releasing control by his neighbor, the experience designer.

Each day we rely on our own trees of knowledge — branches of immeasurable directions and depth, overlapping and crossing one another to form meshed nests of position. The common faith we tend to hold regarding knowledge, is in the strength to overlap our individual trees with one another; the more the overlap, the more the homogenous culture, driving civil movement within this complex ecosystem and jungle we’ve created for ourselves.

Well, some people seem to prescribe to such theories.

In the midst of this information revolution, when we engage in the practice of tagging our information objects, we’re not only engaging in an activity to increase the discovery of our position via the use of common signifiers, we’re implicitly participating in a form of expression — painting our personal mental model of our signified constructs onto the sign itself.

In turn, the degree of shared context an individual holds on the receiving end, determines the degree to which her reception of the sign becomes explicit communication.

Enabled by technology, we can now easily add descriptive tags to the aggregate objects of words, colors, sounds and movement delivered more directly to the branches of each other’s trees. In this flip scenario of retrieval, we now rapidly stumble across these additions, assigning them as variants of welcome or disruptive bits of information.

In any case, our common trees of knowledge are being affected… they are evolving.

To this day, these particular words of Ferdinad de Saussure cannot escape my purview:

In the lives of individuals and of societies, language is a factor of greater importance than any other. For the study of language to remain soley the business of a handful of specialists would be a quite unacceptable state of affairs. In practice, the study of language is of some degree or other the concern of everyone.

photo by heather allison

If Bill hadn’t stepped into the wrong plane at the wrong time in the fall of 1994, he would’ve witnessed rapid advancements of the inner-workings of the web — specifically the participatory meshing of topics, interests, desires and perspectives via individual and social tagging through citizen blogging, vlogging, podcasting, etc.

The post-modern, knowledge craving, subversive side of Bill would be beaming right about now… just about as brightly as the multinational, career for-hire professor.

In the name of knowledge, and a hat-tip to my mentor, I think I’ll be busy late into the evening this October 30th.