About three weeks' ago I responded to an inquiry from New Scientist reporter Duncan Graham-Rowe about HP's semantic blogging project. This post updates my comments to Duncan, my posts from this blog, and contributions to the Pie wiki.  

Steve Cayzer presented Semantic Blogging for Bibliography Management at the 2003 Blogtalk conference. HP labs is running the project as one of several semantic web demonstration pilots. Bloggers get to add research-appropriate references to their blogs, with the data structures preserved at each step in a blogging ecology (blog CMS, UI, syndication, aggregation, newsreader, and presentation).

What's a structure-enahnced blog item?

Packages of structured data are becoming post components.

The virtue of blogs has been their simplicity. Each post only needs one field, and maybe a title and url.

Not everyone is served well by this lowest common denominator. Sometimes you have a burning need for more structure, at least some of the time.

When you know a subject deeply, and your observations or analysis recur, you may be best served by filling in a form. The form will have its own metadata and its own data model.

Consider a school soccer coach. An after-game report typically includes:

• which teams played,
• where and when,
• officials, and
• a list of game events
  • who scored (and when and how),
  • who received penalties (when and for what), etc.).

Wouldn't it be handy for your blogging tool to:

• understand this structure,
• present an editing form,
• render the form in html to your blog, and
• render the post (including the form) to your rss feed?

News aggregators and news readers should be able to:

• Autodiscover an unknown structure.
• Notify the user that a new structure is available.
• Learn the structure, including entry forms, pick list sources, rendering guidance, and default style sheets.
• Make it available when the blogger is ready to write.

So how would you use this? Broadly,

• people get to express themselves and
• blogs start to interoparate with enterprise applications.  

Recipes and Golf Scores:

You should be able to define your own structure. The most common use of Microsoft Excel is making lists of things. No reason blogs can't give similar freedom to define a new package. Build from scratch or on the shoulders of other package definitions. Just for diversity sake:

• What I'm listening to now. (with enough info that a newsreader could find and play that tune, when the package comes in via RSS.)
• Concert review. (3 points for lighting, 2 for sound, 4 for audience involvement, ...)
• Strange things in my referral log. (A real blog)
• Beer reviews. (A real blog)
• Dating reports. (6 stars on manners, 9 on heat, 2 on wardrobe, ...)
• Burning man barter "haves" and "wants"

Interop with enterprise applications:

So I define a "new customer bio" structure. My customer relationship management system writes RSS for me that includes new customer info. Not only can I cite that post in my blog, but:

• my blog can notify the CRM via trackback
• the CRM can take note of the permalink of my post (for CRM users), and
• the CRM can append changes to data I made with my blogging tool ("He's not really the decision maker.").

Along the way...

• information locked inside an enterprise system become visible to our intranet search engine via my blog.
• more useful content finds its way into enterprise systems.
• transactional data takes on context.

If you're building it, what are some of the architectural concerns?

1. Adaptive. Either you're going to create a special purpose blogging tool, something that understands one particular semantic structure, like rsearch citations. Or you're going to create tools that permit ongoing discovery, learning, and adaptation to new structures.
2. Structure Preserving. You should be able to pass along anything. 
3. Include Inline vs. by Reference. Lots of debate over this.

Some real world examples:

• Qlogger
  • Sub-schemas describe activities (golfing, commuting) and reviews (movies, marijuana). You can see how this creates more comparable data (show me all the movie reviews by warbloggers rated 4 out of 5 stars). Coming soon are trend charts so you can see if you golf game is getting better or getting worse, or if you commute times are better on some days.
• JemBlog
  • The Jena Semantic Web Server.
  • evelopment Diary, Notes, Download, Mailing List, SourceForge
• The Lafayette Project (2003 Q2?)
  • From Meg Hourihan's talk at Reboot (20 June 2003)
    • Today, you can review a book on your blog and a review on Amazon. It would be better if you could just tell Amazon about the review on your site. More distributed. It would be cool to link recipes/reviews to Epicurious and collaboratively filter that info (people who cooked this, also cooked this). You get to own your content but connect with others, retain copyright but still participate in your discussion.
  • What is the Lafayette Project? by Nick Denton

What can happen if lots of people use structure-enhanced blogs?

Individual use may look like this...

1. Bakers define recipe formats.
2. They teach them to their weblog tool (defining a form).
3. They post recipes.
4. Posts with recipes are syndicated.
5. Other bloggers notice the new format. By autodiscovery. By reading weblogs with structured content (sports scores in a nice table, for example).
6. They add it to their own blogging tool.
7. Then when they compose a new post, they can pick components to add (Pillsbury Commercial Recipe Format, imdb Movie Review Format (supported by Amazon), HR department's vacation survey).
8. You manage your list of formats, retiring some, promoting others.
9. Formats proliferate through the blogosphere. Some become de facto leaders for a type of content or transaction. Others carve out niches, perhaps associated with a vendor or an industry.

 Collectively, the blogosphere becomes adaptive.

Most people will still do plain old blogging, lucky if they use a title or main link.

Many will occasionally use a structure. Especially as Blog This buttons proliferate. So you can post an SAP invoice to your intranet blog, for example.

Others will find a few formats that tie in closely with a deep interest or passion, or their jobs. A runner's diary. A movie review. A project status report.

Enterprise applications will read components. So information will flow between the modestly structured blogosphere and the highly structured business infrastructure. For example,

1. I'm a Macy's buyer.
2. I get an echo feed from J.D.OracleSoft for orders processed in my category.
3. I post some of them to my team weblog, marking them up with comments about the intended customers, thoughts about the vendor meeting, and terms that didn't fit in the software (front row seats for the family).
4. J.D.OracleSoft reads my feed, looking for changes to the order, and adding the post's permalink to the transaction record.

In the list of working examples, I mentioned Qlogger, the first blogging service that publishes structured blogs. Qlogger today isn't the whole semantic blog picture. While they format blogs according to their structures, (a) they don't publish those structures in a machine readable format like XML, and (b) they don't provide a programmatic way to ask questions of the server using those structures. For example, show me the most popular running shoe by people who jog less than 5 miles.

Back to the juicy stuff at HP labs...

The HP demonstrator requirements are brilliant.

First, they will show that tools, augmented by the semantic web, can create convenience and value for individual participants. In their test, they're helping researchers to cite more and better with less effort.

Second, they show that decentralized ideas and information, created and exposed using the tools of the semantic web, create synergy and and community benefits. It will be easy to see which citations are more popular, to find like-minded researchers based on citations, and to form new communities of practice using this information. By building semantic tools along the edge of the Internet, new patterns of user behavior and information should emerge from the links and content that connect them.

Third, the project will showcase blogs integrated into other information systems. This portends your blogging tools evolving into personal portals. So you'll read not just other blogs but also feeds from your favorite workplace software. We already see blogs that enhance posts before publishing by automatically grabbing related links from Google and appending them to the post. The demonstrator makes that a two way conversation, between bloggers and the existing world of IT services and applications.

Maybe the most valuable contribution will be giving scientists and technologists a gut sense of how the semantic web will feel as a user. After that, it's just engineering and a field day for the social scientists.

Also noteworthy, Easy News Topics by Matt Mower (England) and Paolo Valdemarin (Italy) is a protocol that lets people add ontologic context to blog posts (e.g. this post is about /animals/things-that-live-in-the-water/fish/salmon) while continually improving and shaping topic trees in a distributed way. As each blog grows to hold thousands of posts (hey, a few posts a day adds up), we'll need ways to help us find posts by topic. ENT is a parallel effort to make it easy to share those topic notations.

[a klog apart]

Michael Gartenberg and Chris Sells blame the user, not the tool. Room enough for both, I think. There is always the question of doing the wrong thing more efficiently. Or using a tool as a crutch or substitute for presentation prep and delivery skills.

I'm quite fond of some common tips:

1. One idea per slide
2. Replace words with pictures
3. Fewer words are better. One word or phrase is best.  
4. Talk over slide transitions

That said, nothing replaces rehearsal to perfection, clear organization, ruthless editing, and people skills.

The data is trapped in my TiVo, just waiting to get out. I installed the home network option, so an app on my desktop PC serves mp3 and photos for playing on the TiVo. Well and good.

The problem: TiVoCorp doesn't let your TiVo serve its programming metadata data back to the desktop. While I can program the TiVo from the web to record a program, the TiVo can't tell you that you're overriding previous programming. No access to your season passes or recommendations.

So I'm programming blind.

Another wishlist item: Publish RSS of my viewing history. The better to blog about movies, news, and other shows. I can't wait for TiVo Hacks.

[a klog apart]

In the September 2003 issue, by Halley Suitt.

A tech-savvy employee has something to say about everything at surgical glove manufacturer Lancaster-Webb. When she raved on-line about an older style of gloves, sales unexpectedly shot up. And when she posted damaging information about a potential customer's business practices, the deal collapsed. Is "Glove Girl" a priceless marketing weapon or a grave security risk?

[a klog apart]