Plural processor inferencing system and method

• US 4,982,340 A
• Filed: 06/28/1988
• Issued: 01/01/1991
• Est. Priority Date: 06/30/1987
• Status: Expired due to Term

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Oyanagi, Shigeru, Suzuoka, Takashi, Nakamura, Sadao, Fujita, Sumikazu
• Primary Examiner(s): MACDONALD, ALLEN R
• Application Number:
  US07/212,490
• Time in Patent Office: 917 Days
• Field of Search: 364/513, 364/300, 364/200, 364/133
• US Class Current: 706/55
• CPC Class Codes: G06F 8/45 Exploiting coarse grain parallelism in compilation, i.e. parallelism between groups of instructions
• CPC Class Codes: G06N 5/047 Pattern matching networks; Rete networks

Abstract

An inference processing system comprises a plurality of processors assigned for nodes of a semantic network respectively to carry out parallel processing. Each of the processors stores tables and linking information. The tables are made by arranging rules and prepared for goals respectively. Each table contains the names of links to follow when a certain goal is given and markers for specifying the kinds of the links. In the table, the link names and markers are set as outputs with respect to the given goal as an input. The linking information contains information of links connected to a node for which a corresponding one of the processors is assigned. When the goal and a marker are inputted to any one of the processors, the processor finds an output marker and output links in the tables and linking information and outputs the output marker to the output links, thus achieving an inference process.

Claims

1. An inference processing system for finding solutions to a goal given by an external unit by matching the goal with certain facts expressed in the form of a semantic network composed of nodes and links that connect the nodes to each other, comprising:

a plurality of processors provided for the nodes of the semantic network respectively, each of said processors having means for storing tables prepared for respective goals to be given and for storing linking information related to links connected to a corresponding one of the nodes, each of the tables storing names of output links to follow when at least a goal corresponding to the table is inputted thereto as well as output markers for specifying kinds of the output links; and

connecting means for connecting said processors to each other according to the links of the semantic network,

each of said processors including means for specifying, when a goal and a marker are inputted thereto from the other of said processors through said connecting means, an output link name and an output marker in the tables and for specifying linking information according to the inputted goal and marker, and for sensing the goal and specified output marker to links having the specified output link name, thereby carrying out parallel inference processing.

2. The inference processing system as claimed in claim 1, wherein each of said processors includes means for specifying an output link name and an output marker in the table and for specifying linking information according to the inputted goal when only a goal is inputted from said external unit, and including means for sending the goal and the specified output marker to links having the specified output link name.

3. The inference processing system as claimed in claim 1, wherein the tables stored in each of said processors are specified respectively by the goals to be given thereto, and each of the tables holds the names of input links and input markers to be referred when a goal and a marker are given to the table and holds the names of output links and output markers to be outputted corresponding to the input link names and input markers.

4. The inference processing system as claimed in claim 1, wherein each of said processors includes means for storing tables prepared for respective goals to be given thereto, each of the tables holding the names of output links to follow when at least a goal corresponding to the table is inputted thereto as well as output markers for specifying the positions of the output links to follow with respect to solutions.

5. The inference processing system as claimed in claim 1, wherein said connecting means comprise two-way connecting passages for connecting the processors in parallel.

6. The inference processing system as claimed in claim 1, wherein the linking information stored in each of said processors indicates all the links connected to a node for which a corresponding one of said processors is assigned.

7. The inference processing system as claimed in claim 1, wherein the external unit comprises a host computer.

8. The inference processing system as claimed in claim 1, wherein each of the tables stored in each of said processors includes a set of link names and markers for input and a set of link names and markers for output, the set for input and the set for output of the link names and markers being able to be inverted to realize two-way characteristics.

9. An inference processing system for finding solutions to a goal given by an external unit by matching the goal with certain facts expressed in the form of a semantic network composed of nodes and links that connect the nodes to each other, comprising:

a plurality of processors provided for the node of the semantic network respectively, each of said processors having storage means for storing tables which are identifiable with a goal or a marker from the other of said processors given thereto which include means for storing linking information related to links connected to a corresponding one of the nodes, each of the storage means storing input link names or input markers as well as storing output link names or output markers corresponding to the input link names or the input markers; and

connecting means for connecting sid processors to each other according to the links of the semantic network,

each of said processors having means for specifying, when a goal and a marker are inputted thereto from the other of said processors through said connecting means, an output link name and an output marker in the tables and for specifying linking information according to the inputted goal and marker, and for sending the goal and specified output marker to links having the specified output link name, each of said processors further including means for specifying, when only a goal is inputted thereto from the external unit, an output link name and an output marker in the tables and for outputting the goal and the specified output marker to links having the specified output link name.

10. The inference processing system as claimed in claim 9, wherein said connecting means comprise two-way connecting passages for connecting the processors in parallel.

11. The inference processing system as claimed in claim 9, wherein the linking information stored in each of said processors indicates all the links connected to a node for which a corresponding one of said processors is assigned.

12. The inference processing system as claimed in claim 9, wherein the external unit comprises a host computer.

13. An inference processing method for finding solutions to a goal given by an external unit by matching the goal with certain facts expressed in the form of a semantic network composed of nodes and links that connect the nodes to each other, said method comprising:

storing, within each of a plurality of processors provided for the respective nodes of the semantic network, a table prepared for the respective goals to be given and linking information related to links connected to a corresponding one of the nodes, each of the tables storing names of output links to follow when at least a goal corresponding to the table is inputted thereto as well as output markers for specifying kinds of the output links;

connecting said processors to each other according to the links of the semantic network;

specifying, when a goal and marker are inputted to one of said processors form another of said processors, an output link name and an output marker in the tables and specifying linking information according to the inputted goal and marker; and

then

sending the goal and the specified output marker to links having the specified output link name, thereby carrying out parallel inference processing.

14. The inference processing method as claimed in claim 13, wherein, when only a goal is inputted into a processor from the external unit, an output link name and an output marker are specified in the corresponding table and linking information is specified according to the inputted goal, and the goal and the specified output marker are sent to links having the specified output link names.

15. The inference processing method as claimed in claim 13, wherein each of the tables are specified by the goals to be given thereto and each processor holds the names of input links and input markers to be referred to when a goal and a marker are given to the table and holds the names of output links and output markers to be outputted corresponding to the input links and the input markers.

16. The inference processing method as claimed in claim 13, wherein each of said processors stores tables prepared for respective goals to be given thereto, and each of the tables holds the names and output links to follow when at least a goal corresponding to the table is inputted thereto as well as output markers for specifying the positions of the output links to follow with respect to the solutions.

17. The inference processing method as claimed in claim 13, wherein the linking information stored in each processor indicates all the links connected to a node for which a corresponding one of said processors is assigned.

18. The inference processing method as claimed in claim 13, wherein each of the tables includes a set of link names and markers for input and a set of link names and markers for output, the markers being capable of being inverted to realize two-way characteristics.