Semiotic decision making system used for responding to natural language queries and other purposes and components therefor

• US 6,275,817 B1
• Filed: 12/22/1999
• Issued: 08/14/2001
• Est. Priority Date: 07/30/1999
• Status: Expired due to Term

• Current Assignee: AutoGnomics Corp., Unisys Corporation
• Original Assignee: AutoGnomics Corp., Unisys Corporation
• Inventors: Parker, Charles Austin, Waller, John C., McCrory, Duane J., Reed, Frederick W., Pendergraft, Eugene David
• Primary Examiner(s): Chaki, Kakali
• Application Number:
  US09/469,836
• Time in Patent Office: 601 Days
• Field of Search: 706/45, 706/46, 706/59, 706/61, 704/1, 704/231, 704/241, 704/251, 704/256, 704/257, 707/3, 707/6, 707/102
• US Class Current: 706/45
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/268 Morphological analysis

Abstract

A decision making system uses semiotic processing modules to transform a training corpus of information, in the form of sequential sets of symbols, into a knowledge database. The knowledge database is thereafter used to make decisions relating to queries input in the same type of training corpus symbols. In the knowledge base, the system stores data representations of analyses of subsets of the training corpus sets of sequential elements. The knowledge base data representations comprise predicates and elemental and non-elemental acts. An inductive processor recursively processes the training corpus sets by evaluating the relationship and frequency of occurrence of individual elements and sets of elements in the training corpus. After processing of the training corpus is completed, the resultant knowledge base is used to evaluate queries in a performance mode of operation.

Claims

1. A dyadic semiotic processing module for a semiotic decision making system wherein a training corpus of information in the form of sequential sets of elements, where the number of elements in each set does not exceed a selected finite number, is used to create a database which is thereafter used to make decisions relating to queries input in the same type of elements, comprising:

a knowledge base for storing data representations of analyses of subsets of predefined sets of sequential elements;

said knowledge base data representations comprising predicates and elemental and non-elemental acts wherein;

each predicate is associated with a class of one or more acts such that each act is associated with only one class of acts;

each elemental act represents a set element and defines a single act class of a corresponding elemental predicate; and

each non-elemental act represents a sequence of a case predicate followed by a result predicate, such that all non-elemental acts are recursively defined as representations of one or more sets of sequential elements and each predicate represents the sets of sequential elements represented by each act within its associated class of acts;

an input for receiving sets of lineally-related elements and storing a representation of each element as an elemental act record in said knowledge base;

an inductive processor associated with said knowledge base which recursively processes the received sets by evaluating the relationship and frequency of occurrence of individual elements and sets of elements based upon the lineal relationship of those elements as received by said input to facilitate the generation of predicate records and non-elemental act records based thereon which are then stored in said knowledge base;

said inductive processor generating statistics regarding the correlation of the stored data representations with said predefined sets of sequential elements;

said inductive processor including an associated inductive analysis space memory used for analyzing each of said predefined sets of sequential elements;

said analysis space memory having, for any given set of n elements, data cells for n progressively higher levels of analysis conducted by said inductive processor;

each level of analysis associated with one or more data cells in which are stored predicates and associated acts;

a first level of analysis associated with n cells, each first level cell for receiving one or more elemental predicates and corresponding elemental acts representing one of said n set elements such that all n set elements are associated with one of said first level cells; and

each cell associated with an ith analysis level, where I is an integer from 2 to n, for receiving predicate and associated act representations of analyses of a different sequence of I sequential elements based on the predicates previously stored in lower level cells in correlation with the acts stored in said knowledge base.

2. A dyadic semiotic processing module according to claim 1 wherein:

each act has an associated probability based upon its frequency of use in analysis space cells; and

each ith level cell, for i>

1, includes an associated rank for each predicate within the cell, which rank is based on probabilities of acts recursively associated with the predicate in correlation with the sequence of i elements associated with the cell.

3. A dyadic semiotic processing module according to claim 1, wherein the inductive analysis space further comprises:

records, containing references to predicates and acts in the knowledge base;

records in the inductive analysis space further containing pointers to records representing predicates previously received in lower level cells in correlation with the acts stored in the knowledge base; and

each cell containing pointers to records in the inductive analysis space that represent the predicates received in that cell.

4. A dyadic semiotic processing module according to claim 3, wherein the inductive analysis space further comprises:

predicate records and act records, each predicate record containing a pointer to a predicate in the knowledge base, each act record containing a pointer to an act in the knowledge base;

each act record in the inductive analysis space further containing pointers to predicate records representing predicates previously received in lower level cells in correlation with the acts stored in the knowledge base;

each predicate record in the inductive analysis space further containing pointers to those act records in the inductive analysis space that represent acts in the knowledge base associated with the predicate in the knowledge base to which the predicate record points; and

each cell containing pointers to predicate records in the inductive analysis space that represent the predicates received in that cell.

5. A triadic semiotic processing module for a semiotic decision making s ystem wherein a training corpus of information in the form of sequential sets of symbols is converted into sets of lineally-related elements where each element represents a node of a multi-node tree form having a single highest level node and a plurality n of lowest level nodes where the lowest level nodes represent one or more sequential sets of training corpus symbols, where each lowest level node is associated with a single higher level node and where each node which is not a lowest level node is associated with exactly two lower level nodes whereby the total number of nodes N is equal to 2n−

1, from which a database is created which is thereafter used to make decisions relating to queries input in the same type of training corpus symbols, the module comprising;

a knowledge base for storing data representations of analyses of subsets of predefined sets of lineally-related elements;

said knowledge base data representations comprising predicates and elemental and non-elemental acts wherein;

each predicate is associated with a class of one or more acts such that each act is associated with only one class of acts;

each elemental act represents a set element and defines a single act class of a corresponding elemental predicate; and

each non-elemental act represents a sequence of a case predicate followed by a result predicate, such that all non-elemental acts are recursively defined as representations of one or more subsets of lineally-related elements of a predefined set and each predicate represents the subsets of lineally-related elements represented by each act within its associated class of acts;

an input for receiving sets of lineally-related elements and storing a representation of each element as an elemental act record in said knowledge base;

an inductive processor associated with said knowledge base which recursively processes the received sets by evaluating the relationship and frequency of occurrence of individual elements and sets of elements based upon the lineal relationship of those elements as received by said input to facilitate the generation of predicate records and non-elemental act records based thereon which are then stored in said knowledge base;

said inductive processor generating statistics regarding the correlation of the stored data representations with said predefined sets of lineally-related elements;

said inductive processor including an associated inductive analysis space memory used for analyzing each of said predefined sets of lineally-related elements;

said analysis space memory having, for any given set of N elements, data cells for N progressively higher levels of analysis conducted by said inductive processor;

each level of analysis associated with one or more data cells in which are stored predicates and associated acts, a first level of analysis associated with N cells, each first level cell for receiving one or more elemental predicates and corresponding elemental acts representing one of said N set elements such that all N set elements are associated with one of said first level cells; and

each cell associated with an ith analysis level, where I is an integer from 2 to N, for receiving predicate and associated act representations of analyses of a different subset of I lineally-related elements of the set of N elements based on the predicates previously received in lower level cells in correlation with the acts stored in said knowledge base.

6. A triadic semiotic processing module according to claim 5 wherein:

each act has an associated probability based upon its frequency of use in analysis space cells; and

each ith level cell, for i>

1, includes an associated rank for each predicate within the cell, which rank is based on probabilities of acts recursively associated with the predicate in correlation with the subset of i lineally-related elements associated with the cell.

7. A triadic semiotic processing module according to claim 5, wherein the inductive analysis space further comprises:

records, containing references to predicates and in the knowledge base;

records in the inductive analysis space further containing pointers to predicate records representing predicates previously received in lower level cells in correlation with the acts stored in the knowledge base; and

each cell containing pointers to predicate records in the inductive analysis space that represent the predicates received in that cell.

8. A triadic semiotic processing module according to claim 7, wherein the inductive analysis space further comprises:

predicate records and act records, each predicate record containing a pointer to a predicate in the knowledge base, each act record containing a pointer to an act in the knowledge base;

each act record in the inductive analysis space further containing pointers to predicate records representing predicates previously received in lower level cells in correlation with the acts stored in the knowledge base;

each predicate record in the inductive analysis space further containing pointers to those act records in the inductive analysis space that represent acts in the knowledge base associated with the predicate in the knowledge base to which the predicate record points; and

each cell containing pointers to predicate records in the inductive analysis space that represent the predicates received in that cell.