WWW2007: Prabhakar Raghavan on “Web N.0: What sciences will it take?”

Note: These are rough notes from the WWW2007 conference.

What sciences get us to the next stage of the Web?

There is content in various forms: editorial (newspapers), free (stream of consciousness), and commercial.  There is an audience that consumes all the different flavors of content.  They also enhance the content.  They help to filter out the content.  They purchase the content.

In the middle, you have the corporations. All the corporations are built in two pursuits, we build the audience and then we monetize them.

We face the challenge of search.  There is the algorithmic results and there is the paid monetized results.

The premise is that people don’t want to search.  People just want to get tasks done.  We spend a lot of time searching but computers spend half a second solving your problems.

Search has become task centric.  Its about understanding your intent.  When you searching for Papa Johns, are you looking to invest or are you looking for a pizza?

The grand challenge is to build general platforms for task centric needs.  Generality is critical.

Community computing…

For User Generated Content, there is 5-10 Gb created/day.  There is also user generated metadata, like anchor text, tags, page views, and reviews.  We need to start taking advantage of non-anchor text meta data.

START Meta Data

  • Star – bookmarking
  • Tags – label for retrieval
  • Access – viewing a page
  • Routing
  • Text

Within Flickr, you have a community of photo users.  People tag their photos.  People tag other people’s photos.  This is the wisdom of the crowds helping with search.

How do you use these tags better?  How do you cope with spam?  Whats the rating and reputation system?   What are the incentive mechanisms?

ESP game is a great example of incentivizing tagging.  People feel like they’re winning something.  At the same time, there is meta data that can be harvested.

There are a lot of good QnA systems.

What assignment of incentives leads to good user behavior? What’s good user behavior?  Whom do you trust and why?

Online media experiences…

We’re building audiences and monetizing them.  Yahoo! is a media company that is built on technology.  They’re there to serve media to users.

We need a science of online audience engagement.  It’s not just people interacting with their computers.  Its people interacting with other people.

We need to understand things like:

  • Why do people lurk or participate?
  • Why do people create new online personas?
  • Why are YouTube, MySpace, and Flickr successful?
  • What new genres that are emerging and what can we provoke?

There are different dimensions of experience:

  • duration – short to long
  • ephemerally – forgotten to remembered
  • social context – alone to with others

Second Life is a long duration, remembered, and something you do with other people.

Audience engagement is measured by the number of page view.   With ajax, page views wasn’t as good.  People were just staying on the same page.  They weren’t refreshing.

What are other ways to measure user engagement?  There are different levels. We need metrics for each of those levels.

Microeconomics meets CS…

On the right of most search results you’ll see a ranked list of advertisements.  Classically, slots that are higher up will get more clicks.

Three problems

  • Match ads to query/context
  • Order the ads
  • Pricing on a click-through

The first two have to do with IR.  The last two also have to with economics.  For ordering,  GoTo/Overture used ordering of the ads by bid.   What has become popular is revenue ordering.  What gets clicked on a lot is what goes to the top and what doesn’t get clicked a lot falls to the bottom.

Monetization and economic value are an intrinsic part of the system.

Advertisements

One thought on “WWW2007: Prabhakar Raghavan on “Web N.0: What sciences will it take?””

  1. Great post! The issue of intent and motivation is part of the black box of webbased activity and interaction. The next web app, widget, or search engine that taps into that will have quite a useful and profitable commodity on their hands. Do you think http://www.url.com, which seems like Digg for search engines is a step in the right direction. Would people take the 5 seconds (lets call it the google gut-check) to enter their intent on an opt-in basis? would self reporting fail? would the “thieves” or “hoodlums” and otherwise nafarious characters overwhelm the wisdom of crowds?

    You mentioned Google abandoning products earlier…it would be too bad if they let this one slip through their fingers at whatever stage of the game.

    Hope the conference is going well!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s