Granted Patent B2

US 12,346,318 · App. 18/593,760 · Granted Jul 1, 2025

Inventor: James Kaplan (Marina Del Rey, CA)

Assignee: MEETKAI, INC.

G06F16/2433

Abstract

Claims (41)

1. A method for automatically generating question and query pair from a knowledge base comprising the steps of:

accessing a knowledge base;

extracting, via processing module stored in a memory and executed by a processor, one or more entities and one or more predicates from said knowledge base;

generating, via a baseline template generator module, a baseline question template from a context free grammar rule;

inputting the one or more entities and the one or more predicates into the baseline question template to determine a numbering of the one or more predicates;

generating a numbered question template based on the numbering;

generating a typed question template by traversing the numbered question template through a predicate type-graph;

generating a protocol and RDF query language (SPARQL) query by inputting the one or more entities and the one or more predicates into a SPARQL query template;

generating a question by inputting the one or more entities and the one or more predicates into the typed question template;

constructing a question and query pair by combining the question and the SPARQL query; and

storing the question and query pair in the memory for future access.

2. The method of claim 1 wherein the knowledge base is an open knowledge base.

3. The method of claim 2 wherein the open knowledge base is WikiData.

4. The method of claim 1 wherein the SPARQL query comprises one or more of the following types: single-entity, multi-entity, count, two intention, ranking, and filtering.

5. The method of claim 1 wherein the predicate-type graph comprises at least two predicates connected by an edge.

6. The method of claim 5 wherein the edge may be uni-directional or bi-directional.

7. The method of claim 5 wherein the edge comprises of legal paths.

8. A system for automatically generating question and query pair from a knowledge base comprising:

a memory, a processor, and non-transitory machine readable code stored in the memory and executed by the processor, the non-transitory machine readable code comprising:

a knowledge processing module configured to access a knowledge base and identify an entity and two predicates from the knowledge base;

a baseline template generator configured to receive input of a context free grammar rule and to generate a baseline question template based on the input of the context free grammar rule;

a numbering module configured to number the two predicates, and to generate a numbered question template based on the baseline question template and the numbering of the two predicates;

a traverse predicate module configured to traverse the numbered question template through a predicate type-graph and to generate a typed question template based on the traverse;

a protocol and RDF query language (SPARQL) query module configured to generate a SPARQL query based on input of the entity and the predicate; and

a question and query constructor module configured to generate a question based on input of the entity and the predicate to the typed question template, and to generate a question and query pair using the question and the SPARQL query to then store the question and query pair in the memory for future lookup.

9. The system of claim 8 wherein the memory is stored in, and the processor operates on, one or more of the following: a user device, a server, a cloud.

10. The system of claim 8 wherein the numbering comprises appending a nesting order to the two predicates.

11. The system of claim 10 wherein the SPARQL query module uses the same nesting order.

12. The system of claim 8 wherein the knowledge processing module configured to identify two entities, and the numbering module is further configured to number the two entities.

13. The system of claim 8 further comprising a database comprising of one or more question and query pairs.

14. A system to automatically generate questions comprising:

a memory, a processor, and non-transitory machine readable code stored in the memory and executed by the processor, the non-transitory machine readable code comprising:

a baseline template generator configured generate a baseline question template from a context free grammar rule;

a numbering module configured to input one or more entities and one or more predicates, append a nesting order to the one or more predicates, and to generate a numbered question template based on the baseline question template and the nesting order;

a traverse predicate module configured to generate a typed question template based on the traverse of the numbered question template through a predicate type-graph; and

a question constructor module configured to generate a question based the predicate to the typed question template, the one or more entities, and the one or more predicates;

wherein the generated question is stored in the memory.

15. The system of claim 14 wherein the one or more entities and the one or more predicates are identified from a knowledge base.

16. The system of claim 14 wherein the predicate type-graph comprises two nodes and one edge, the node comprising a predicate, and the edge comprising the ontological type of the predicate.

17. The system of claim 16 wherein the predicate type-graph comprises a first edge and a second edge, and wherein each node is treated as a function from the first edge to the second edge, or from the second edge to the first edge.

18. The system of claim 16 wherein the edge is directional or bi-directional.

Assignments (1)

Continuity (3)

References Cited (90)