Method for processing an expert system rulebase on a system having limited memory

• US 4,809,219 A
• Filed: 06/26/1985
• Issued: 02/28/1989
• Est. Priority Date: 06/26/1985
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Iwaskiw, Christine T., Flagg, Richard L., Ashford, Thomas J., Starbird, Roberta P., Burns, Nancy A.
• Primary Examiner(s): Shaw, Gareth D.
• Assistant Examiner(s): CHUN, DEBRA
• Application Number:
  US06/749,076
• Time in Patent Office: 1,343 Days
• Field of Search: 364/200 MS File, 364/900 MS File
• US Class Current: 706/50
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

A method for running an expert system on a data processing system in which the Rulebase is segmented into contextual units so that the system memory can accommodate each unit. Portions of the Rulebase containing data or knowledge that is not needed in a particular application may then be eliminated. The segmenting of the Rulebase also allows the contextual units of the application to be paged into and out of the system as needed. The various intersegment relationships are managed by the system as units are paged into and out of memory. A second Rulebase unit may be called by the first unit and executed at any time during the processing of the first Rulebase. Provision is made in the form of a Global attribute to selectively identify and store the data that is relevant to the application and that has been accumulated up to the point where another unit is called so that at some time later in the process, when the system returns to the first unit, it can proceed from the last RULE node that was processed. Selected data that has been identified as global during processing of the first unit is used to update selected nodes in the second unit of the Rulebase after it has been paged into the system. Selected data collected during the processing of the second unit is passed to the first unit when the system returns to complete processing that unit.

Claims

1. In a method of tracing a Rulebase of an expert system with a data processing system having a processing unit and a program which functions as an inference engine, a main memory unit which a memory capacity which is less than the size of the Rulebase and a storage unit having a storage media with a storage capacity capable of storing said entire Rulebase and connected to said processing unit for selectively transferring data there between, the improvement comprising the following steps in combination;

(1) segmenting said Rulebase into a plurality of contextual units each of which has a size less than said size of said main memory unit, each said contextual unit having a plurality of Rule trees having a Goal node at its root and a plurality of other nodes at the leaves of said tree, each said node of said Rulebase comprising a data record having a plurality of predefined fields, each said record including at least one field which is interrelated to an identical field in at least one other said record in said Rulebase, each set of said identical related fields defining a different linked list of interrelated records, with each of said sets including a different group of said records, said segmenting step including the step of,

(A) inserting a Rulebase Call action in a preselected field of selected said records of said Rulebase to limit each said contextual unit to a size that can be accommodated by said main memory,

(2) transferring a first contextual unit from said media to said main memory,

(3) tracing in a predetermined order a plurality of said Rule trees in said first unit with said inference engine with each said tree being traced in a back chaining traversal of the nodes of said tree,

(4) interrupting said step of tracing in response to said inference engine detecting said RULEBASE CALL during the processing of said preselected field of one of said selected said records,

(5) transferring a second contextual unit to said main memory from said media in response to said detected RULEBASE CALL,

(6) transferring said first contextual unit to said storage media when the amount of said memory available for said second unit being called is inadequate simultaneously to store said first and second units, and

(7) updating selected nodes in said called second unit with data collected during said step of tracing of said first contextual unit prior to said interruption.

2. The method recited in claim 1 further including the steps of tracing said second contextual unit until each Goal node is concluded, and exchanging said second contextual unit with said first contextual unit in said main memory unit to permit said first contextual unit to be traced from the point of said interruption to another RULEBASE CALL or until each Goal node in said first contextual unit has been concluded.

3. The method recited in claim 2 further including the step of updating said first contextual unit with data collected during the said step of tracing said second contextual unit after said step of exchanging said contextual units.

4. The method recited in claim 3 in which each said step of tracing said Rule trees of said first and second units includes the steps of processing each of the said other nodes at the leaves of said trees to obtain data to assist in arriving at a conclusion for the Goal node of said tree.

5. The method recited in claim 4 in which said steps of processig each of said other nodes includes the step of processing at least one Evidence type node or at least one External type node.

6. The method recited in claim 5 further including the step of assigning a Global attribute to a first set of nodes located in different rulebase units and said steps of updating said first and second units causes the data collected for one node of said set during the processing of one of said rulebase units to be transferred to another node of said set in a different rulebase unit.

7. The method recited in claim 6 further including the step of developing a Global list of each of said node in said first contextual rulebase unit that has been assigned a Global attribute as said first contextual unit is transferred into said memory and transferring the data collected by said node assigned a Global attribute to said Global list prior to processing said second contextual rulebase unit.

8. The method recited in claim 7 in which said steps of updating further include the step of tranferring from said Global list any data collected by a node in said first contextual rulebase unit that has a name identical to a node in said second contextual rulebase unit that is also assigned a global attribute.

9. The method recited in claim 8 further including the step of updating at the conclusion of processing a node for the first time in one contextual rulebase unit any other node in the same said one contextual rulebase unit which has an identical name.

10. The method recited in claim 9 in which said step of updating a node for the first time includes the further step of updating at least one Evidence type node.

11. The method recited in claim 10 further including the step of providing a terminal for displaying data in said system to an operator and a keyboard for entering data into said system, and said step of tracing a plurality of said Rule trees in said first unit includes the step of displaying on said terminal, a class question when said evidence node is processed.

12. The method recited in claim 11 further including the step of comparing the data entered from said keyboard by said operator in response to said question with a predetermined set of allowable responses.

13. The method recited in claim 12 in which said step of tracing also includes the step of displaying data to the operator which requires said operator to only acknowledge the receipt of said data.

14. The method recited in claim 13 in which said step of displaying data to said operator includes the step of displaying a predetermined directive to said operator which said operator is expected to follow and further including the step of said operator pressing a predetermined key on said keyboard to acknowledge completion of said directive.

15. The method recited in claim 4 further including the step of maintaining a Concluded Goals list of Goal nodes that have been concluded during the processing of said Rulebase unit.

16. The method recited in claim 15 in which said step of maintaining said Goals list includes establishing one record for each Goal node that was concluded, in which said one record comprises one field for an identification of said Goal node and another field for the conclusion data.

17. The method recited in claim 16 further including the step of processing said Concluded Goals List after all said units have been processed to present to said operator a predetermined number of conclusions and a numerical value indicating the relative correctness of each of conclusions.

18. The invention recited in claim 4 in which said step of tracing includes tracing all Goal trees in said first contextual units so that each goal is concluded before said first unit is considered exhausted.