Semantic network machine for artificial intelligence computer

• US 4,815,005 A
• Filed: 11/19/1987
• Issued: 03/21/1989
• Est. Priority Date: 11/29/1986
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Oyanagi, Shigeru, Nakamura, Sadao, Fujita, Sumikazu
• Primary Examiner(s): Pellinen, A. D.
• Assistant Examiner(s): WILLIAMS, HOWARD L
• Application Number:
  US07/122,446
• Time in Patent Office: 488 Days
• Field of Search: 364/513, 364/300
• US Class Current: 706/55
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation

Abstract

A semantic network machine is applied to an artificial intelligence computer for performing inferential retrieval with respect to a knowledge base. A main associative memory stores the knowledge base consisting of knowledge data arranged to form a semantic network. Each knowledge data consists of a set of an object, an attribute, and a value. A sub associative memory is connected in parallel with the main memory and stores specific knowledge data including "is-a" attributes of the knowledge data. In an inferential retrieval mode, when a question associated with a given object is input, an initial retrieval condition for retrieving, from the knowledge base, knowledge data necessary for answering the question is defined. While the main memory is accessed using the initial retrieval condition, the sub memory is also accessed simultaneously to read out data representing another object concept associated with the object concept included in the question by the "is-a" attribute from the sub memory in a parallel manner. The readout data is temporarily stored in a buffer memory. When retrieval in the main memory fails, the initial retrieval condition is updated using the data stored in the buffer memory to generate a secondary initial retrieval condition supplied to the main memory, which is then successively accessed using the updated condition.

Claims

1. A semantic network machine comprising:

first associative memory means for storing a plurality of knowledge data arranged to form a semantic network, thereby constituting a knowledge base, each of the knowledge data having a set of an object, an attribute, and a value, and the attribute including an "is-a" attribute;

second associative memory means, arranged in parallel with said first associative memory means, for storing a plurality of specific knowledge data including the "is-a" attributes of the plurality of knowledge data; and

retrieval control means, connected to said first and second associative memory means, for parallel-accessing said second associative memory means while accessing said first associative memory means so as to retrieve knowledge data necessary for answering an input question in an inferential retrieval mode, for, when the accessing in said first associative memory means fails, feeding back an access result, which represents another object concept associated with an object concept included in the input question by the "is-a" attribute, from said second associative memory means to said first associative memory means, and for successively accessing said first associative memory means using the access result as a part of a new retrieval condition.

2. The machine according to claim 1, further comprising:

buffer memory means, connected to said second associative memory means, for temporarily storing data representing the access result from said second associative memory means.

3. The machine according to claim 2, wherein said retrieval control means controls said buffer memory means to feed back the data stored therein to said first and second associative memory means, so that a retrieval condition for said second associative memory means is updated using the data.

4. The machine according to claim 3, wherein said first associative memory means comprises a main associative memory, and said second associative memory means comprises a sub associative memory.

5. A semantic network machine which is applied to an artificial intelligence computer performing inferential retrieval of a knowledge base, said machine comprising:

main associative memory means for storing a plurality of knowledge data arranged to form a semantic network, thereby constituting a knowledge base, each of the knowledge data having a set of an object, an attribute, and a value, and the attribute including an "is-a" attribute;

sub associative memory means, arranged in parallel with said main associative memory means, for storing a plurality of specific knowledge data including the "is-a" attributes of the plurality of knowledge data;

buffer memory means, connected to said main and sub associative memory means, for temporarily storing data read out from said sub associative memory means; and

retrieval control means, connected to said main and sub associative memory means, for, (i) in an inferential retrieval mode, defining, for said main associative memory means, a first initial retrieval condition necessary for retrieving, from said knowledge base, knowledge data necessary for answering an input question associated with a given object, (ii) for defining, for said sub associative memory means, a second initial retrieval condition necessary for retrieving the given object included in the question and another object having the "is-a" attribute from the storage content of said sub associative memory means, (iii) for accessing said sub associative memory means using the second initial retrieval condition simultaneously with accessing of said main associative memory means using the first initial retrieval condition so as to read out data representing another object concept associated with the object concept included in the question by the "is-a" attribute from said sub associative memory means in a parallel manner and to store the readout data in said buffer memory means, (iv) for, when the retrieval in said main associative memory means fails, updating the first initial retrieval condition using the data stored in said buffer memory means to form a secondary first retrieval condition supplied to said main associative memory means, and v) for successively accessing said main associative memory means using the secondary first retrieval condition.

6. The machine according to claim 5, wherein said retrieval control means controls said buffer memory means and said sub associative memory means so that the data stored in said buffer memory means is fed back not only to said main associative memory means but also to said sub associative memory means to update the second initial retrieval condition for said sub associative memory means using the data, thereby successively accessing said sub associative memory means.

7. The machine according to claim 6, wherein said sub associative memory means stores the specific knowledge data and data representing objects and values excluding attributes in a one-to-one correspondence manner.

8. The machine according to claim 7, wherein when a plurality of knowledge data are extracted as a result of retrieval in said sub associative memory means, said buffer memory means stores these knowledge data at the same time.

9. The machine according to claim 8, wherein said buffer memory comprises a queuing buffer memory.

10. A method of performing inferential retrieval with respect to a knowledge base, said knowledge base being constituted by a plurality of knowledge data which are arranged to constitute a semantic network and each of which has a set of an object, an attribute, and a value, and being stored in a main associative memory, said method comprising the steps of:

(a) storing a plurality of knowledge data including "is-a" attributes of the plurality of knowledge data in a sub associative memory;

(b) defining, for said main associative memory, a first initial retrieval condition necessary for retrieving, from said knowledge base, knowledge data necessary for answering an input question associated with a given object, in an inferential retrieval mode;

(c) defining, for said sub associative memory, a second initial retrieval condition for retrieving the object included in the question and another object having the "is-a" attribute from the storage content of said sub associative memory;

(d) accessing said sub associative memory using the second initial retrieval condition simultaneously with accessing of said main associative memory using the first initial retrieval condition and reading out data representing another object concept associated with the object concept included in the question by the "is-a" attribute from said sub associative memory;

(e) temporarily storing the data in a buffer memory;

(f) when the retrieval in said main associative memory fails, updating the first initial retrieval condition using the data stored in said buffer memory and forming a secondary first retrieval condition supplied to said main associative memory; and

(g) successively accessing said main associative memory using the secondary first retrieval condition.

11. The method according to claim 10, wherein while said main associative memory is successively accessed, the second initial retrieval condition is updated using the data stored in said buffer memory to form a secondary second retrieval condition supplied to said sub associative memory, with which the retrieval in said sub associative memory is performed in a parallel manner.