Rule acquisition for expert systems

• US 4,816,994 A
• Filed: 12/04/1984
• Issued: 03/28/1989
• Est. Priority Date: 12/04/1984
• Status: Expired due to Fees

• Current Assignee: Tektronix Incorporated (Fortive Corp.)
• Original Assignee: Tektronix Incorporated (Fortive Corp.)
• Inventors: Alexander, James H., Phillips, Brian, Messick, Steven L., Freiling, Michael J.
• Primary Examiner(s): Shaw, Gareth D.
• Assistant Examiner(s): Kriess, Kevin A.
• Application Number:
  US06/678,193
• Time in Patent Office: 1,575 Days
• Field of Search: 364/200 MS File, 364/900 MS File, 364/513, 364/513.5, 364/300
• US Class Current: 706/59
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: Y10S 706/916 Electronic or computer, internal or network, circuit

Abstract

A rule acquisition method for expert systems receives input from the expert, one word at a time, and checks each word by determining alternative parses for the partial sentence as so far entered. Feedback is provided to the expert when a word does not conform to a required grammar, this feedback comprising a menu of legal next choices. When the sentence is completed it is translated into the form of a diagnositc rule. An expert system for troubleshooting electronic equipment is disclosed employing rules generated in the aforementioned manner.

Claims

1. The method of acquiring rules for a computer-based expert system, said rules being characterized by conditions involving queries having more than one possible answer, said method comprising,

serially receiving portions of a sentence input from an expert,

checking said portions as they are received for their confirmation to a predetermined grammar based upon the logical formation of rules for a given subject matter,

providing sentence portion selection information to said expert, wherein said selection information is given to the expert by way of feedback in instances where the expert inputs a sentence portion that does not conform to said predetermined grammar, and

translating the sentence into the form of a diagnostic rule when the sentence is complete.

2. The method according to claim 1 wherein said feedback is provided in menu-fashion of possible next choices as would conform to said predetermined grammar.

3. The method of acquiring diagnostic rules for a computer-based expert system, said method comprising,

serially receiving the words of a sentence input of an expert,

checking each word as it is entered by simultaneously determining alternative parses for the partial sentence so far entered,

providing feedback information to the expert when the word entered does not match any of the current partial parses, and

translating the sentence into the form of a diagnostic rule.

4. The method according to claim 3 wherein said feedback information is provided in the form of a menu of possible next sentence portions for selection by the expert.

5. The method according to claim 3 wherein the development of said parses is implemented with a breadth-first parsing algorithm on a semantic grammar.

6. The method of acquiring rules for a computer-based expert system, said rules being characterized by conditions involving queries having more than one possible answer, said method comprising,

receiving a first sentence portion as automatically determined to conform to a predetermined allowable grammatical construction fitting within an allowable logical sequence,

also receiving at least a second sentence portion as automatically determined to conform to a predetermined allowable grammatical construction fitting within an allowable logical sequence with the first sentence portion, and

employing said conforming portions to provide an expert rule,

wherein at least upon non-conformation of a sentence portion, feedback is provided in the form of a choice of acceptable inputs as would satisfy currently allowable grammatical construction.

7. The method according to claim 6 wherein said choice is provided in the form of a selectable menu.

8. The method according to claim 6 wherein the last mentioned allowable grammatical construction is logically generated in sequence upon the first mentioned allowable grammatical construction.

9. The method of acquiring rules for a computer-based expert system, said rules being characterized by presentation of queries having more than one answer, said method comprising,

serially receiving portions of a sentence from an expert wherein each of said portions conforms to a grammar based upon the logical formation of rules for said expert system and wherein a said sentence portion is received as a choice among alternatives electronically presented to said expert wherein successive menu-selections are provided to form succesive sentence portions, as prior menu-selections are completed, and

employing said conforming portions to provide an expert rule.

10. The method of acquiring rules for a computer-based expert system, said rules being characterized by a predetermined grammar and conditions involving queries having more than one possible answer, said method comprising,

serially receiving portions of a sentence input from an expert,

in response to an already received portion of sentence input, generating alternatives for a next portion of a sentence permitted by said predetermined grammar,

accepting a further portion of said sentence input from said expert as matches one of said alternatives,

as said further sentence portion is accepted, continuing to generate allowable alternatives and accept matching input for following sentence portions from said expert until a statement is completed according to said predetermined grammar, and

transforming the completed statement into the form of a diagnostic rule.

11. The method according to claim 10 including graphically presenting said alternatives for a next portion of said sentence to said expert and receiving said further portion of said sentence input from said expert in the form of a selection by said expert of one of the alternatives presented.

12. The method according to claim 10 including graphically presenting alternatives for a next portion of said sentence to said expert in response to input by said expert not matching one of said alternatives.

13. The method according to claim 10 including receiving numerical input from said expert as one or more of said sentence portions.

14. The method according to claim 10, wherein transforming the completed statement into the form of a diagnostic rule comprises:

constructing functional description equations from fragments of said statement,

combining functional equations to form functional structures, and

converting said functional structures to rule clauses.