Inference system

• US 4,704,695 A
• Filed: 05/14/1985
• Issued: 11/03/1987
• Est. Priority Date: 06/26/1984
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Kimura, Kazuhiro, Ishida, Katsuyo
• Primary Examiner(s): RUGGIERO, JOSEPH F
• Application Number:
  US06/733,721
• Time in Patent Office: 903 Days
• Field of Search: 364/513, 364/148, 364/194, 364/200 MS File, 364/900 MS File, 364/513.5, 364/300, 382/14, 382/15
• US Class Current: 706/50
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

In an inference system in accordance with the present invention, an attributive information is supplemented to a knowledge unit for the purpose of smooth execution of inference, and a knowledge base is constructed with this pair as the building unit. In connection with this, functions that are in response to the purpose, such as selection or rejection of a knowledge unit and alteration in the order of search, are added to the inference organ, based on the attributive information. Employed as an attributive information is, for example, the information on the frequency of use of each knowledge unit. The inference organ renews the information on the frequency of use for every use of each knowledge unit. In addition, a housing space is allocated to each knowledge unit based on the information on the frequency of use. This means, for example, that a knowledge unit with higher frequency of use is housed in an area of the inference organ such as core memory that may be accessed with ease.

Claims

1. A method of inferring a suitable answer to an interrogation utilizing a plurality of knowledge units in accordance with corresponding attributive information, which comprises the steps of:

(a) accessing the knowledge units and corresponding attributive information from knowledge accumulation means, said corresponding attributive information including frequency-of-use information;

(b) rearranging the accessed knowledge units and corresponding attributive information into a frequency-of-use order based on the frequency-of-use information;

(c) storing the rearranged knowledge units and corresponding attributive information in work domains of a memory;

(d) searching the knowledge units stored in step (c) in accordance with the frequency-of-use information successively from knowledge units having a greater frequency-of-use to knowledge units having a lesser frequency-of-use;

(e) selecting the suitable answer based on one of

(1) knowledge units stored in the work domains, and

(2) knowledge units generated by inference of syllogism;

(f) updating the frequency-of-use information of the selected knowledge units.

2. The method as claimed in claim 1, wherein said frequency-of-use order is from the knowledge units having the greater frequency-of-use to the knowledge units having the lesser frequency-of-use.

3. The method as claimed in claim 1, further including the step of comparing frequency-of-use information of the knowledge units with each other so that the knowledge units are rearranged in decreasing order from greatest frequency-of-use to lesser frequency-of-use.

4. The method as claimed in claim 1, wherein the knowledge units having greater frequency-of-use are stored in a storage zone of one work domain capable of faster access than a storage zone of another work domain which stores the knowledge units having lesser frequency-of-use.

5. An inference system for inferring a suitable answer to an interrogation utilizing a plurality of knowledge units in accordance with corresponding attributive information, comprising:

(a) input means for inputting an inference instruction signal;

(b) knowledge accumulation means for accumulating the plurality of knowledge units together with their corresponding attributive information, said corresponding attributive information including frequency-of-use information;

(c) work domains for temporarily storing the knowledge units and their corresponding attributive information in accordance with the frequency-of-use information;

(d) knowledge housing and managing means for reading the knowledge units and corresponding attributive information from the knowledge accumulation means, for rearranging the knowledge units and corresponding attributive information into a frequency-of-use order based on the frequency-of-use information, and for storing the knowledge units and corresponding attributive information in said work domains successively in accordance with the frequency-of-use order;

(e) inference processing means for selecting the suitable answer based on one of

(2) knowledge units stored in said work domains, and

(2) knowledge units generated by inference of syllogism based on the inference instruction signal; and

(f) output means for displaying the selected knowledge units.

6. The inference system as claimed in claim 5, wherein each corresponding attributive information includes category information, such as an inference rule, fact, and goal or target, an object, a field of use, and users.

7. The inference system as claimed in claim 6, wherein said knowledge housing and managing means updates the frequency-of-use information of the selected knowledge units.

8. The inference system as claimed in claim 6, wherein said work domains comprise:

at least one storage zone of one work domain for storing knowledge units having a greater frequency-of-use,

at least one storage zone of another work domain for storing knowledge units having a lesser frequency-of-use.

wherein said at least one storage zone of said one work domain has a faster access than said at least one storage zone of said another work domain.