The world's information is moving to the Web, and people are spending more and more time online. To help people deal with the vast quantities of information available on the Web, a variety of tools have emerged. Search engines return ranked lists of Web pages; browser history tools remember what pages you have visited; social bookmarking services such as del.icio.us and digg let you share Web pages with others.
However, all these tools assumed that the Web was a collection of static documents, and that the information you were searching for was embedded in one of those documents. Yet this is no longer the case. The Web is changing from a collection of documents to a platform for applications. And increasingly, it's how you interact with those applications that's important. Let me give you an example.
I recently served as treasurer for the IUI 2009 conference on intelligent user interfaces. One of my duties as treasurer was to submit the proposed conference budget to the ACM for approval. The process for budget submission included filling out a multi-page Web form that asked for various bits of information about the conference: the full name of the conference, conference description, expected attendance, how many papers would be published, our paper acceptance criteria, how many pages per paper, how many workshops and tutorials there would be, etc.
It took me several days to fill out that form. I collected information from the conference Web site, from the program chairs, and several other places in order to assemble all the information I needed. Once I was finished, I breathed a sigh of relief and moved on to other tasks. I never expected to see that form again.
However, several months later, I was contacted by the incoming conference chairs for the 2010 conference. "Do you remember how the budget approval process works?" they asked. "Could you help our treasurer with that process this year?"
Hmm. Well, I could have sent a link to the Web page that kicks off the submission process. I could have used a search engine, or my browser history, to find the Web form again, and sent them the URL. However, almost all the information that would be useful to the incoming treasurer was contained in how I answered all the questions on the form. Most of those answers don't change from year to year; if they did, it's must easier to start from last year's information and update it than to start from scratch. The information they wanted was not contained within a single Web page, it was embodied in how I interacted with those pages.
What if we had Web tools that tracked not just what Web pages we visited, but what we did on those pages? This idea is the basis of our Smarter Web initiative, a collection of projects that leverage knowledge about user activity on Web pages to deliver a variety of useful tools. As part of the Smarter Web, we are building a platform for recording and playing back user actions on Web sites, storing them in a natural language representation, and saving sequences of these actions--scripts--in a centralized repository.
The flagship Smarter Web project is CoScripter, the one that got us all started thinking about how to improve the Web experience. CoScripter lets you record how to do a Web task by demonstration, creates a script describing what you did, and saves that to a central server where it is shared with the community to create a user-contributed knowledge base of how to do things on the web. You can download CoScripter today and try it out for free.
Another Smarter Web project is Highlight, a system that lets you create mobile versions of Web sites based on how you interact with the full site on your desktop. For any popular site, everyone uses it differently; for example, on an airline's Web site, some people might be interested in checking in for their flight, while others might want to find low fares to their favorite destination. Highlight leverages traces of a user interacting with the site on their desktop computer to identify which portions of the Web site are needed to complete their task, and it clips just those parts to create a lightweight, task-driven version of the Web site for use on a mobile device.
There are many other examples of Smarter Web projects; our group's recent work includes projects that suggest task completions for blind users, build mashups by demonstration, and collaboratively browse the web over instant messaging. All these projects share a common platform for capturing and reusing the actions people take on Web pages.
What else could we do if we knew what users do on the Web? I think we are opening up a huge opportunity to mine this interaction data and build more systems that help users manage the vast amounts of information that we produce and consume. Does this sound interesting? Let's collaborate!