Kristian J. Hammond, Co-Director, Intelligent Information Laboratory, Department of Computer Science, Northwestern University

You have incredibly valuable content but nobody can tell the difference between it and something that someone has hacked up and thrown on the Internet. People get what they can from it and go away, driving up the bounce rate. Why should anyone in the world get it for money? People can wait it out, and it will be free. NYT will say, "Just take it, at least show up some time."

People think that they can defeat Google on its own turf with search. Like a new soda trying to go head-to-head with Coca-Cola. Unless there is a serious cultural change, the last battle that you fought is going to be fought over and over again. There's an arms race going on, someone out there is trying to bust into your information. We'll crawl the Web and have a parallel information repository to make "one-stop" shopping for our vertical. Nice idea, but still head-to-head with Google. Or, make a deal with Google, but that's just getting your information out more efficiently, doesn't address monetization.

"We suggest leapfrogging the problems altogether..." In cellular telephony, adoption in developing nations surged where land lines were mediocre at best. Leapfrogged rather than trying to make technology work in old environments. We work on frictionless information systems, or intelligent information systems. "Whatever you're thinking, we'll get to your heart's desire." We want to get rid of the search box, so that you don't interact with it directly. Need information systems that are aware of the context of your behavior. Knowing your customer at this second and using it as a driver of information. We're a lab, we don't care where information comes from.

Examples: the desktop relevance engine and "beyond broadcast," for online video, and "make my page," for machine-generated content. Pulls information on the Web based on the document being read (COMMENT: not really so different from "more like this" feature or training data for semantic searches). Really on-point, finds docs first on own Web site, then other sites. Also pulls up discussions, experts, methods, learning, Web and desktop collections. (COMMENT: good idea. Not done often enough). Integrated with desktop activity (COMMENT: like Watson, kind of), can pull from any searchable source. Kind of like doing enterprise portal services for the world).

Then looked at server-side versions of this. Chicago Sun-Times example, other blogs, videos, Web, other S-T news, S-T blogs (COMMENT: This is a lot like Sphere in a sense, all fine and good, but Sphere is not really helping publishers to generate engagement or revenues to any significant degree). In video, main context is watching television. Build a small piece of code for TiVo box, "beyond broadcast," watches television with you, notices what channel you're on, look at closed caption text, feed that information to server, hijack a TiVo remote button, would build a microsite based on the content and the context. Different for cooking show, how-to show, dramas, etc. (COMMENT: This has strong potential as a concept, contextual content and ads for television). Were elated with this, people will come to us, but then YouTube happened and convergence happened anyway without television. Took same system, moved it online. Works pretty much the same, change banner ads based on what' being discussed in the video.

Looked at a few examples, "good eats" on the Food Network, links to Wikipedia entry on corn dogs. Look at other version of recipes, etc. (COMMENT: Again, not much new on one level, but doing it in real-time is a fantastic tool, equivalent to AdWords on Google search). Give them everything that they might be interested in (COMMENT: But do publishers really want more distractions from their content? I agree with the "real-time portal" based on content of interest, but again, Sphere experience raises questions).

"Make my page" function, pulls together new document, primarily links to other document, open to social media editing. Highly relevant, will percolate up to the top and stay there. (COMMENT: Now, that works better. Make the associations of content more enduring, even when they are assembled in real-time. If the aggregation works for one person, it may work for others, and it will be improved for others. Similar to my old concept of distributable objects that grow in value as they are passed from person to person contextually). "We will get you to your heart's desire."

Great stuff, I like the idea of the persistent contextual objects that grow over time, that's the real winning ingredient. Many of the other ideas have been done before, but if you can create content that becomes more popular over time automatically as it expands and transforms over time through contextual use, it will be self-optimizing for engagement.

Labels: enterprise, media, publishing, SIIA Information Industry Summit 2009