Task-based Information Management

Abstract

In a rapidly changing environment such as crisis battle planning or intelligence analysis, a major obstacle to eective collaboration is keeping collaborators informed. Collaborators must have access to all information critical to the performance of their tasks, as soon as that information is created. At the same time, they must not be overloaded with irrelevant information; i.e., they should not need to waste large amounts of time sorting through documents or messages for pertinent information.

Managing information under these conditions presents problems for current information technology, because of the dynamic nature of most collaborative processes. New information is continually created, so a standard query-return model of information retrieval will not suce: team members cannot continually query a database for information they need, especially if they do not know that the information exists. On the other hand, systems that perform information ltering [Belkin and Croft 1992; Yan and Garcia-Molina 1995], where information is routed to users based on user-supplied proles or long-standing queries, are also inadequate. In a dynamic collaborative environment, team members are constantly changing roles within the collaboration, completing some tasks and beginning new ones, while new participants are entering the process as others drop out. A piece of information that is relevant to a team member at one time may not be relevant to that same team member at another time. To deal with this problem, a tool for information management must not only be able to use a team member’s tasks in distributing, retrieving, and storing information, but must also be able to adapt to ongoing changes in those tasks.

We are constructing a system, called the Task-based Information Distribution Environment (TIDE), that delivers information to participants in a dynamic collaboration by evaluating the relevance of incoming and newly generated information to collaborators’ current tasks. This relevance detection process consists of discovering a mapping between two representations|a feature vector description of incoming documents, and task models that represent all collaborators’ current tasks. This paper discusses three current products of our ongoing research effort:

1. an initial formalism for representing analysis tasks and the relationship between those tasks,
2. a method for deriving keyword vector queries from instantiations of the task representation, and
3. a prototype system that delivers documents to users based on their task descriptions.