Inference processing method and apparatus

Claim 2

A high speed inference method in an inference engine comprising a computer (H) for executing programs, a storage (G) for storing a knowledge base including fact knowledge and rules, and a program for compiling the knowledge base to a machine language instruction set for said computer, wherein the fact is converted (Fig. 10) to a machine language instruction set executable by said computer.

Claim 3

A high speed inference method according to Claim 1 wherein the machine language instruction set of the fact has a function for holding values of the fact or addresses of said values in predetermined registers.

Claim 4

A high speed inference method according to Claim 1 wherein the machine language instruction set of the fact has at least a function for holding values of the fact or addresses of said values in predetermined registers, and the machine language instruction set of the rules has at least a function of referring said predetermined registers for matching.

Claim 7

A method for operating an inference engine according to Claim 4 further comprising: a step of generating a control instruction set for executing the condition parts of the rules other than said rule at the end of the calculation of all instantiations satisfying the condition part of said rule or when there is no fact satisfying the condition part of said rule; and a step of generating a fact control table having information for specifying the machine language instruc­tion set, address information of rules which refer the attribute value of the fact and internal number informa­tion for identifying the fact.

Claim 8

A compile type knowledge processing tool comprising: a computer (H) for executing a stored program; a storage (G) for storing a knowledge base including fact knowledge and decision knowledge or rule; said rule including a condition part for specifying a condition applicable to the rule and an execution part which directs execution when the rule is applied; said fact knowledge including attributes of any values; said rules including a rule having a condition part for representing a relation of the attributes and an execution part for modifying a value of the attribute; a program (I1, I4; I2, I3) for converting the knowledge base to a machine language instruction set for said computer; a work area for storing said knowledge base and supporting backtrack for said rule; means (G3) for converting to a machine language instruction set having a function for generating in an area for holding instantiation having at least said rule and the fact knowledge which meets the condition part of said rule; means (Fig. 12) for converting to a control instruction set for executing the condition parts of the rules other than said rule at the end of calculation of all instantiations which meet the condition part of said rule or when there is no fact which meets the condition part of said rule; means for converting the execution part (Fig. 15) of the rule which directs the modification of the attribute value of the fact to a machine language instruction set having a function for modifying the attribute value of the fact and a machine language instruction set (fr, cut) having a function for deleting all instantiations having said fact as elements from the instantiation holding area; means (Fig. 10) for converting to a machine language instruction set for holding the attribute values of the fact in predetermined registers; and means (Fig. 11) for generating a fact management table having information for specifying the machine language instruction set, address information of the rules which refer the fact and internal number information for identifying the fact.

Claim 9

A compile type knowledge processing tool comprising: a computer (H) for executing a stored program; a storage (G) for storing a knowledge base including fact knowledge and decision knowledge or rule; said rule including a condition part for specifying a condition applicable to the rule and an execution part which directs execution when the rule is applied; said fact knowledge including attributes of any values; said rules including a rule having a condition part for representing a relation of the attributes and an execution part for modifying a value of the attribute; a program (I1, I4; I2, I3) for converting the knowledge base to a machine language instruction set for said computer; a work area for storing said knowledge base and supporting backtrack for said rule; means (G3) for converting to a machine language instruction set having a function for generating in an area for holding instantiation having at least said rule and the fact knowledge which meets the condition part of said rule; means (Fig. 12) for converting to a control instruction set for executing the condition parts of the rules other than said rule at the end of calculation of all instantiations which meet the condition part of said rule or when there is no fact which meets the condition part of said rule; means for converting the execution part (Fig. 15) of the rule which directs the modification of the attribute value of the fact to a machine language instruction set having a function for modifying the attribute value of the fact and a machine language instruction set (fr, cut) having a function for deleting all instantiations having said fact as elements from the instantiation holding area; means (Fig. 10) for converting to a machine language instruction set for holding the attribute values of the fact in predetermined registers; and means (Fig. 11) for generating a fact management table having information for specifying the machine language instruction set, address information of the rules which refer the fact and internal number information for identifying the fact.