IBM Research has been active in virtually all areas of web-related research since the inception of the World Wide Web. This has involved publishing high volumes of dynamic data at some of the most highly accessed web sites in the world, devising new techniques for searching and characterizing the web, contributing to standards affecting the entire industry, creating the infrastructure needed for the web to move to XML, developing cooperative and collaborative applications enabled by the World Wide Web, and web access from mobile devices.

Web Performance

We have pioneered new techniques for efficiently serving dynamic content. Our work has addressed the challenging problem of publishing large amounts of dynamically changing content while handling high peak loads. The techniques resulting from our research have been deployed at some of the most highly accessed web sites in the world, including the Sydney 2000 Olympics web site. Over the last few years, these sites have maintained 100 percent availability despite increases of orders of magnitude in publishing volumes and hit rates.

Other web performance work has focused on caching, load balancing, web server acceleration, web traffic characterization, optimization and benchmarking of web server performance, transparent discovery of network performance, and server support for gigabit networking.

XML Infrastructure

We are creating the infrastructure for the web to move to XML as its data exchange format and for much of its data persistence. Several directions are being explored including: a modular XHTML system that will integrate SMIL, XForms, MathML, and P3P into HTML; ability to separate information content and information rendering, and put them together again using XSL; and ability for application or industry-specific markup vocabularies to be described with XML Schema Language and queried with XML Query Language.

Collaboration and Awareness

We have developed a tool for high-level awareness and collaboration, called Livemaps, for projecting live information onto a web site map. We are also investigating how to overcome two of the most difficult aspects of remote collaboration - common ground and group focus. Our work involves modeling group behavior in collaborative environments and using these models to inform design of improved collaboration spaces. Another project focuses on the "distance" among people in a collaborative environment - with the goal of reducing the sense of isolation and providing group awareness while, at the same time, maintaining privacy. In our approach, awareness is provided by multidevice event perception, and distance is adjusted through intelligent multiagent cooperation and negotiation.

Information Retrieval

Another effort is developing techniques to deal with information overload in today's interconnected world of web and intranet servers. Most of us respond to information overload by reading only relevant and authoritative matter. Relevance can be characterized by the documents that the user has seen or liked as well as their link structure. Authority or quality can be attributed to documents, based on hyperlink citations. Various techniques based on machine learning and graph algorithms are being used to mine documents in large hypertext databases for relevance and quality.

We are currently extending these methods to web image retrieval. By analyzing the page-to-image as well as page-to-page link structure, we are able to retrieve relevant images based on text queries. Additionally, we can locate image containers and image hubs, which are defined as web pages that are rich in relevant images, or from which many images are readily accessible.

Web Access from Mobile Devices

We are investigating issues dealing with mobile knowledge seekers who typically need to access information from the web when they are away from their desktops. The constraints imposed by mobile devices with bandwidth and form-factor limitations often make information discovery tasks impractical. We have developed a new approach specifically oriented for the mode of work and the constraints dictated by mobile devices. It combines focused search within specific topics, with encapsulation of topic-specific information in a persistent repository. The repositories are based on "knowledge-agent bases" that comprise all the information necessary to access data about a topic and assist in the full search process, from query formulation assistance to result scanning on the device itself.

.Please contact Paridhi Verma to obtain copies of the Computer Science Brochure