Objective
The main goals of this project are :
1. to specify an open architecture for reasoning and
2. to construct a prototype query generator for literature retrieval.
The term "open" means a system in which the knowledge representation schemata are explicitly defined, and open to user alteration. Similar requirements hold true for the reasoning procedures. An additional requirement is that there are powerful general reasoning primitives to construct the high level queries.
The relation oriented inference system (ROIS) is a knowledge based system (KBS) which makes possible the construction of decision support systems tailored to specific applications. This KBS comprises 3 layers; a basic low level inference engine, a subshell and the knowledge base. The function of the subshell is to provide, on the one hand, a data model for representing (scientific) knowledge and, on the other hand, a high level reasoning model. The significance of the subshell is that reasoning algorithms can be separated from the knowledge representation. As the structure for knowledge representation and the reasoning mechanisms are explicitly defined in the knowledge base, a very open and flexible system is fixed. Using this structure it is possible to define several reasoning strategies and a consequence of the system design is that the expert can build the knowledge base independently.
The system has the following characteristics:
there are self definable reasoning strategies;
there is no need for a knowledge engineer;
there is an effective reduction of search space.
As a means to arrive at a set of requirements for the system specifications, the project will investigate a small scale application using an available tool called MEDES. This application is concerned with making a meaningful connection between patient records and literature. It is felt by the partners that the requirements for such an application together with the existing tool concepts cover many of the demands for a general system. One of the partners, Elsevier, will provide the information and facilities to connect the patient and literature systems syntactically as well as semantically. The outcome is expected to have a great impact on future literature retrieval tools.
MEDES, the results of the implementation, and the experience of the partners will be input to the specification for a system to build in the main phase of AIM.
MEDES is a general tool for representing various types of (deep) knowlege . MEDES supports a representation that emphasizes several types of binary relations and a reasoning language that specifies operations on those relations. Prototype knowledge bases for different ways of exploiting medical KBS have been constructed within MEDES.
This project will explore the extension of MEDES to handle patient records and medical literature - call these medical documents. The types of relations and reasoning which are needed to support storage and retrieval of medical documents will be investigated. Evaluations will be done with real medical knowledge, real patient records, and real medical literature. Within the time-frame of this project, a rapid small-scale implementation in MEDES will be done so as to facilitate verification to the robustness of the extensions.
Main Deliverables :
Specification for an open architecture for (medical) reasoning system. Prototype system for generating queries to a literature retrieval system using patient databases.
Fields of science
Topic(s)
Data not availableCall for proposal
Data not availableFunding Scheme
Data not availableCoordinator
Netherlands