Projects > Watson


Despite the extraordinary range of information made accessible by modern computer systems, users face many obstacles that frustrate their efforts to get the information they need in the mass of data available to them. In many cases, users have not clearly defined the information they are seeking and are unsure of whether or where it exists. Users often find it difficult to formulate effective search queries, resulting in their sifting through an ever-growing pile of useless search results. Watson assists in solving these problems by providing a proactive information retrieval tool designed to bridge the gap between a user's information needs and current public and private information systems.

Watson reads what the user is reading and writing in desktop applications, including Microsoft Word, PowerPoint, Outlook, Internet Explorer and Firefox, and automatically searches for relevant on-line information based on context of the task that the user in engaged in. Without requiring the user's active participation, Watson provides a unified and integrated view on a wide variety of disparate information sources which need not themselves be unified or integrated, and provides an infrastructure for easily adding new data sources to the system. Additionally, Watson provides a way to connect people not just to information, but to other people, based on their information needs. Watson dramatically increases the effectiveness and efficiency of its users, enabling companies to keep pace in a world with infinite information resources.

Watson is built using a flexible, open architecture to facilitate its use with a variety of applications and information sources. Watson is connected to applications such as Microsoft Word using application adaptors that track user actions and access the contents of documents being inspected or manipulated. The application adaptors connect to content analyzers, which use heuristic and statistical techniques to create representations of documents. Representations are shipped to query producers, which build queries aimed at different information sources. Finally, the queries are shipped to information sources via information adapters, and the results are filtered, combined and ordered for presentation to the user via the Watson interface. The entire process often takes only seconds.

Watson's usefulness spans multiple markets and industries, favoring organizations which utilize significant volumes of information. Large organizations drawing from intranet systems, proprietary databases or other external data sources will realize savings in time, effort and money through the efficiencies inherent in the Watson technology.

Project Papers

  • Context Transformations for Just-in-time Retrieval: Adapting the Watson System to User Needs