Abstract:
The user interacts with the information retrieval system through
interface agents by formulating queries. Queries are formulated
in keywords or natural language expressions. Keywords do not
express the domain and task contexts and natural language
queries processing are still inefficient. Query formulation is a
very deteminant factor of the efficiency of the whole system. A
method to take into account the task/domain context query
formulation problem is discussed. The system through a feedback
loop learns the user's profile and builds a set of ad hoc search
agents ad hoc search agents as user customized path. This method
has been tested in Hydraulics, Pneumatics and Seals web
robot (HPSBot) a multi agents information retrieval system.
HPSBot is a corporate project initiated by the State Key
Laboratory of Fluid Power to fill the gap between the increasing
corporate information need and the state of the information
retrieval technology.
Key words: Feedback Query formulation User profile Query refining Expected utility
*
This project
is supported by National Natural Science Foundation of China (59975079)
|