Blog

Or search support forum

What's Global Moxie?

Global Moxie specializes in mobile design strategy and user experience for a multiscreen world. We offer consulting services, training, and product-invention workshops to help creative organizations build tapworthy mobile apps and effective websites. We're based in Brooklyn, NY. Learn more.

On Shelves

Books by Josh Clark

Tapworthy: Designing Great iPhone Apps

Best iPhone Apps: The Guide for Discriminating Downloaders

iWork ’09: The Mising Manual

Moxiemail

Enter your e-mail to receive occasional updates:

Lightning! Blimps! Submarines! And, um, Machine Tags!

Posted Jun 18, 2007 (updated Jul 6, 2007)

Hack Day - Bridge Logo

Lightning struck (twice!). Rain followed. Submarines mobilized. Tomatoes flew. Yahoo knitted. Dr. Who descended. Werewolves prowled. Rabbits flagged semaphore. Blimps flew (kind of). And friends, much code was hacked.

Well. That was awesome. I’m bleary-eyed but totally energized after London’s Hack Day. Over 500 developers turned up for the 24-hour hackfest, and in the end, over 70 projects were presented.

I was particularly charmed by Ann McMeekin’s “buzz knit” hack to generate knitting patterns from the content of Yahoo news feeds (“maybe one day I’ll knit my blog”). And like pretty much everyone in the room, I was floored by Steffan Jones’ prize-winning Mac dashboard gadget to show you a slideshow of photos related to the songs playing in iTunes. (Steffan, when are you going to release that baby?)

Hack Day - Rain
Hacking in the rain. Photo by Josh Russell.

The event was thoroughly engaging and altogether humbling. The amount of know-how, creativity and sheer geekery in the room was overwhelming, a potent reminder of how much there is left to learn. It’s a feeling that provokes both anxiety and inspiration. Mostly inspiration. It’s just plain exciting to be part of a profession and community whose frontiers are expanding so fast. So much to explore.

And yeah, the frontier spirit was very much in evidence. Organizers and participants good-naturedly weathered an unlikely collection of setbacks. We battled the elements for an hour or two after a freak lightning strike wigged out the fire system and opened giant vents in the ceiling, letting the rain into a very electrified venue. (“Uh, does anyone here know anything about fire prevention computer systems?” asked event co-organizer Matt Cashmore). What’s a visit to London without a little rain? And, oh yeah, the network was down for several hours.

Through it all, though, there was a palpable sense of enthusiasm and generosity that set it apart from any conference I’ve attended. The London geek community is a fun crowd, affable hosts to those of us visiting from countries across Europe.

Look what I made

Hack Day - Presentation
Presenting my hack. Photo by cackhanded.

Mine was a modest hack, a little web app that generates machine tags to describe any text content that you throw at it.

Machine tags work just like regular tags, except that they’re specially formatted to include data that can be understood by software. This means that they tend to be much more descriptive and accurate than traditional tags. You can use them to associate data with a specific map location, airplane model, just about anything.

So my idea was to create a machine-tag format based on Wikipedia topics, allowing any content creator to tag content with any topic in Wikipedia. By using Wikipedia as an index, this format provides very specific identification of content across a vast knowledge domain. Call it the Dewey Decimal System for the web: “The Wiki Decimal System.”

In general, the problem with machine tags is how to make them easy to add for regular folks. Although the format itself is simple, the tags are typically lengthy and require you to know the data ID for what you want to tag.

Enter my hack: A web page that takes your text and builds the list of Wikipedia machine tags automatically.

A card catalog

Hack Day - "The Wiki Decimal System"
“The Wiki Decimal System” on the big screen. Photo by cackhanded.

As it turned out, figuring out and building the tags was a cinch. The challenge was figuring out a good interface for displaying and editing them. In the end, though, I was really proud of the elegance of the final interface, definitely the most interesting part of the project. I wish I’d shown off the interface a bit more in my presentation, but I sped through it when panic set in at the approach of my 90-second limit. Whaddya gonna do?

Anyhoo... the tag generator finds all of the phrases that seem important in your text and displays each phrase as a library catalog card. Click the tab of the card to view it (and again to return it to the stack).

Each card contains three links to Wikipedia pages related to that phrase. If any of those pages are not important to your text, click the link’s delete button to remove it. If the entire phrase isn’t relevant, you can likewise remove the entire card from the stack. The list of corresponding machine tags is updated as you make these changes.

Want more tags? Just submit new text, and more cards are added to the stack.

Give it a try yourself: The Wikipedia Machine Tag Generator is online here.

And there are photos of my 90-second presentation here, here, here, here, here, here.

June 19 update: I just added a quick explanation of both the app and the tag format, linked from the web app page. I'll do a more formal explanation of the project later, but this should at least provide a basic orientation for newcomers.

What’s next: Machine tags and me

I had a really pleasant chat with Chris Sizemore of the BBC and Rachel Lovinger of Razorish after the presentations. It seems they’ve also been thinking about how to tag content using Wikipedia. We’re all interested in promoting adoption of this (or a similar) tag format, and I’m hoping we’ll explore it further.

My weekend hack might also be a good starting point as a prototype for how to make it as painless as possible for content creators to machine-tag their content.

Meantime, I’d also like to make Big Medium’s nascent tagging infrastructure a bit more machine-tag friendly with options to hide/display machine tags. In most cases, for example, you don’t want your machine tags to show up in your tag cloud. Likewise, I’ll be thinking through other interesting/useful interfaces to make it easy to get machine tags into your Big Medium content; geo-tagging seems especially useful.

Whatever I wind up doing with machine tags, I had a great time building this little gizmo over the weekend. Hack Day was a terrific event with a genuinely warm spirit and an inspiring level of brain power. I learned a lot, I met interesting folks, and I’m already looking forward to next year, assuming the organizers are willing to brave the lightning again.

Tags: , , , ,

Want more? Recent blog entries...

Comments

5 comment(s) on this page. Add your own comment below.

Jun 19, 2007 6:10am [ 1 ]

That's a nice write up and a nice app Josh. I'm not sure how well the concept works, but you can only see that if you try it, and that's what you did, very nicely and very quickly.

What's not obvious from the UI is how the phrases extracted from the text relate to the tags. Also the layout seems to suggest that you go from the text to the tags to the phrases, whereas actually you go from the text to the phrases to the tags. I have a suggestion. Why not show a smaller list of tags and show the phrase that it derives from in the list. Something like:

phrase :: tag

In addition, why not just show the top search hit. Then users can use the phrase's tab to change the tag (or ignore it or add more tags if they like) from the phrase's wikipedia search result list. Also if you took this route, you could show more results per phrase (up to seven maybe) in the tab.

Josh
Jun 19, 2007 8:07am [ 2 ]

Matt, many thanks for the great, useful feedback. I think you're right that the layout should be tweaked to flow more obviously left-to-right or top-down to visually display the process better: text leads to phrases, phrases lead to wikipedia topics, topics lead to tags.

You put your finger on the challenge of the UI: I struggled with sorting out how many tags and phrases to display, particularly since the "best" Wikipedia topic for each phrase wasn't always the top result. In order to be sure that the most relevant tags made it into the tag list, I just dumped all of them in there. But I agree that this means too many tags, often on tangential topics and not visually tied to their original context.

For this particular app, simplicity should win out: Better to present fewer tags, but with the opportunity to dive deeper into a larger pool of results. So I think your suggestion to use just one tag per "phrase card" is a good one. Visitors can remove any "bad" topics from the phrase card until the most relevant topic is the top result for that phrase. Since you pretty much have to review all of the cards for quality review in the current layout, the revised interface wouldn't make for more work -- probably less, in fact.

Thanks for the fresh eyes; I'll take a look at incorporating these suggestions over the next week or so.

Mar 15, 2008 2:27pm [ 3 ]

josh, how are you?

i finally had time to put my prototype together. it uses a different approach than you used, but the intent is very similar. i think you are onto something...

http://sells.welcomebackstage.com:5000/item/submit

http://www.slideshare.net/guest2c797e/wikipedia-as-controlled-vocabulary/

all the best!

Oct 17, 2009 3:46pm [ 4 ]

Like the generator. Looks handy :) As for your mention of wanting to handle hide/display machine tags, I currently do that on my blog. All of my blog posts say "Tagged With > tag1 tag2". If you click on the arrow, the tags appear as the true machine tags they are.

Also, if you're looking for a way to organize machine tagged content without overwhelming users, you may want to try machine tag trees. I use them to display my machine tagged posts.

Feb 4, 2011 7:11pm [ 5 ]

Pretty nifty hack!

Take a peek at the Wikipedia Concept Extractor which could amplify the quality of the tags:

http://sourceforge.net/projects/conceptextract/

Add a Comment

Don't be shy.

(Use Markdown for formatting.)

This question helps prevent spam:

Download Big Medium
Try it free for 30 days, or buy to unlock.

Blown Away

“I’m blown away by Josh Clark’s deep understanding of the iPhone user experience.”
—Jürgen Schweizer, founder of Cultured Code, maker of Things iPhone app

“It’s rare to find a person like Josh Clark who speaks so intently to the topic of interface design and mobile devices.”
—John Maeda, president of Rhode Island School of Design

“If you have time to read only one book on what makes apps successful, it is Tapworthy by Josh Clark.”
—Andreas Sjostrom, manager of mobile solutions, Sogeti