Granted Patent B2

US 12,475,113 · App. 17/457,445 · Granted Nov 18, 2025

Inventors: Jing Li (Beijing, CN); Jian Wang (Beijing, CN); Jian Min Jiang (Beijing, CN); Zi Ming Huang (Beijing, CN); Zhen Zhang (Beijing, CN); Wanqing Liu (Hefei, CN)

Assignee: International Business Machines Corporation

G06F16/24522G06F16/243G06F40/295G06N5/02

Abstract

Claims (56)

1 . A computer implemented method comprising:

obtaining, by one or more processing units, a pair including a question and an answer in natural language, the pair originating from an online resource;

determining, by one or more processing units, at least one entity in the question and an entity type of each of the at least one entity in consistence with a schema of a knowledge graph;

identifying, by one or more processing units, a subset of candidate query templates based on the entity type of each of the at least one entity, wherein the candidate query templates are generated based on the schema of the knowledge graph and traverse a schema space of the schema, wherein the schema space is represented by a graph of nodes representing entity types of entities in the knowledge graph, and wherein the graph is traversed into a sub-graph by a query statement and the sub-graph is constrained by a maximal length of a path;

composing, by one or more processing units, a set of queries by populating the at least one entity into each of the subset of candidate query templates;

executing, by one or more processing units, the set of queries on the knowledge graph to generate respective answers;

identifying, by one or more processing units, a first answer from the respective answers that is matching with the answer in the pair; and

determining, by one or more processing units, a candidate query template, from the subset of candidate query templates, corresponding to the first answer as a query template.

2 . The computer implemented method of claim 1 , wherein the candidate query templates comply with Cypher, and wherein the candidate query templates are generated by using a MATCH clause of a Cypher query statement.

3 . The computer implemented method of claim 2 , wherein a directed edge linking one node with another node of the nodes represents a relation of the one node with the another node.

4 . The computer implemented method of claim 3 , wherein using the MATCH clause of Cypher query statement comprises:

using, by one or more processing units, the MATCH clause of Cypher query statement to traverse the sub-graph of the graph, and wherein the sub-graph complies with constraints limiting a size of the sub-graph.

5 . The computer implemented method of claim 1 , further comprising:

including, by one or more processing units, the candidate query template determined as the query template in a library of query templates.

6 . The computer implemented method of claim 1 , wherein executing the set of queries on the knowledge graph to generate the respective answers comprises:

composing, by one or more processing units, the respective answers in natural language based on data retrieved from the knowledge graph in response to execution of the set of queries on the knowledge graph.

7 . The computer implemented method of claim 1 , further comprising:

determining, by one or more processing units, a relationship of an entity type with at least one other object in the question, and

wherein identifying the subset of candidate query templates based on the entity type of each of the at least one entity further comprises:

identifying, by one or more processing units, a subset of candidate query templates based on the relationship of the entity type with the at least one other object in the question.

8 . A system, comprising:

a processor; and

a memory containing a program that, when executed on the processor, performs an operation, the operation comprising:

obtaining a pair including a question and an answer in natural language, the pair originating from an online resource;

determining, by one or more processing units, at least one entity in the question and an entity type of each of the at least one entity in consistence with a schema of a knowledge graph;

identifying a subset of candidate query templates based on the entity type of each of the at least one entity, wherein the candidate query templates are generated based on the schema of the knowledge graph and traverse a schema space of the schema, wherein the schema space is represented by a graph of nodes representing entity types of entities in the knowledge graph, and wherein the graph is traversed into a sub-graph by a query statement and the sub-graph is constrained by a maximal length of a path;

composing a set of queries by populating the at least one entity into each of the subset of candidate query templates;

executing the set of queries on the knowledge graph to generate respective answers;

identifying a first answer from the respective answers that is matching with the answer in the pair; and

determining a candidate query template, from the subset of candidate query templates, corresponding to the first answer as a query template.

9 . The system of claim 8 , wherein the candidate query templates comply with Cypher, and wherein the candidate query templates are generated by using MATCH clause of Cypher query statement.

10 . The system of claim 9 , wherein a directed edge linking one node with another node of the nodes represents a relation of the one node with the another node.

11 . The system of claim 10 , wherein using the MATCH clause of Cypher query statement comprises using the a MATCH clause of a Cypher query statement to traverse the sub-graph of the graph, and wherein the sub-graph complies with constraints limiting a size of the sub-graph.

12 . The system of claim 8 , further comprising:

including the candidate query template determined as the query template in a library of query templates.

13 . The system of claim 8 , wherein executing the set of queries on the knowledge graph to generate the respective answers comprises:

composing the respective answers in natural language based on data retrieved from the knowledge graph in response to execution of the set of queries on the knowledge graph.

14 . The system of claim 8 , further comprising:

determining a relationship of an entity type with at least one other object in the question, and wherein identifying the subset of candidate query templates based on the entity type of each of the at least one entity further comprising:

identifying a subset of candidate query templates based on the relationship of the entity type with the at least one other object in the question.

15 . A computer program product comprising:

a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to perform an operation, the operation comprising:

obtaining a pair including a question and an answer in natural language, the pair originating from an online resource;

determining, by one or more processing units, at least one entity in the question and an entity type of each of the at least one entity in consistence with a schema of a knowledge graph;

identifying a subset of candidate query templates based on the entity type of each of the at least one entity, wherein the candidate query templates are generated based on the schema of the knowledge graph and traverse a schema space of the schema, wherein the schema space is represented by a graph of nodes representing entity types of entities in the knowledge graph, and wherein the graph is traversed into a sub-graph by a query statement and the sub-graph is constrained by a maximal length of a path;

composing a set of queries by populating the at least one entity into each of the subset of candidate query templates;

executing the set of queries on the knowledge graph to generate respective answers;

identifying a first answer from the respective answers that is matching with the answer in the pair; and

determining a candidate query template, from the subset of candidate query templates, corresponding to the first answer as a query template.

16 . The computer program product of claim 15 , wherein the candidate query templates comply with Cypher, and wherein the candidate query templates are generated by using a MATCH clause of a Cypher query statement.

17 . The computer program product of claim 16 , wherein a directed edge linking one node with another node of the nodes represents a relation of the one node with the another node.

18 . The computer program product of claim 17 , wherein using the MATCH clause of Cypher query statement comprises using the MATCH clause of Cypher query statement to traverse the sub-graph of the graph, and wherein the sub-graph complies with constraints limiting a size of the sub-graph.

19 . The computer program product of claim 15 , further comprising:

including the candidate query template determined as the query template in a library of query templates.

20 . The computer program product of claim 15 , wherein executing the set of queries on the knowledge graph to generate respective answers comprises:

composing the respective answers in natural language based on data retrieved from the knowledge graph in response to execution of the set of queries on the knowledge graph.

Assignments (1)

Continuity (1)

References Cited (34)