Granted Patent B2

US 12,326,871 · App. 18/507,769 · Granted Jun 10, 2025

Inventors: Karina Elayne Kervin (Sacramento, CA); Jian Wu (Round Rock, TX); Sibasis Das (Kolkata, IN); Radha Mohan De (Howrah, IN); Swaminathan Balasubramanian (Troy, MI); Cheranellore Vasudevan (Bastrop, TX)

Assignee: INTERNATIONAL BUSINESS MACHINES CORPORATION

G06F16/254G06F16/215G06F16/2246G06N5/022

Abstract

Claims (35)

1. A computer implemented method of detecting duplicate tables comprising:

converting syntactical data from relational tables from a schema of a database to knowledge graphs having semantic data;

mapping nodes in the knowledge graphs to sources in the relational tables;

applying graph matching to the knowledge graphs;

detecting semantically equivalent tables from the relational tables by assessing a degree of matching between matched knowledge graphs based on equivalent data from common linked-pair nodes, defined from relationships, from the knowledge graphs; and

generating a corrective flag to eliminate the semantically equivalent tables from the relational tables from the schema of the database based on the degree of matching.

2. The computer implemented method of claim 1 , wherein the relational tables comprise data arranged in columns.

3. The computer implemented method of claim 1 , wherein the duplicate tables are present in a cloud based data lake.

4. The computer implemented method of claim 1 , wherein the graph matching comprises using at least one of domain specific ontologies, metadata of the nodes, metadata of entities, metadata of fields, and graph manipulation algorithms to determine matches in graphs.

5. The computer implemented method of claim 4 , wherein graph matching by domain specific ontologies includes separating the matched knowledge graphs into table-based sub-graphs and look for similarity between sub-graphs.

6. The computer implemented method of claim 1 , wherein graph matching is a full match between at least two knowledge graphs.

7. The computer implemented method of claim 1 , wherein graph matching comprises a partial match between at least two knowledge graphs.

8. The computer implemented method of claim 1 , wherein graph matching comprises an algorithm selected from the group consisting of graph edit distance, graph kernel, graph embedded for general vector representation, graph neural networks, and combinations thereof.

9. A system for detecting duplicate tables including a hardware processor; and a memory that stores a computer program product, the computer program product of the system includes instructions comprising:

convert, using a hardware processor, syntactical data from relational tables from a schema of a database to knowledge graphs having semantic data;

map, using a hardware processor, nodes in the knowledge graphs to sources in the relational tables;

apply, using a hardware processor, graph matching to the knowledge graphs; and

detect, using the hardware processor, semantically equivalent tables from the relational tables by assessing; a degree of matching between matched knowledge graphs based on equivalent data from common linked-pair nodes, defined from relationships, from the knowledge graphs; and

generate, using the hardware processor, a corrective flag to eliminate the semantically equivalent tables from the relational tables from the schema of the database based on the degree of matching.

10. The system of claim 9 , wherein the relational tables comprise data arranged in columns.

11. The system of claim 9 , wherein the graph matching comprises using at least one of domain specific ontologies, metadata of the nodes, metadata of entities, metadata of fields, and graph manipulation algorithms to determine matches in graphs.

12. The system of claim 11 , wherein graph matching by domain specific ontologies includes separating the matched knowledge graphs into table-based sub-graphs and look for similarity between sub-graphs.

13. The system of claim 9 , wherein graph matching is a full match between at least two knowledge graphs.

14. The system of claim 9 wherein graph matching comprises a partial match between at least two knowledge graphs.

15. The system of claim 9 , wherein graph matching comprises an algorithm selected from the group consisting of graph edit distance, graph kernel, graph embedded for general vector representation, graph neural networks, and combinations thereof.

16. The system of claim 9 , wherein the duplicate tables are present in a cloud based data lake.

17. A computer program product for detecting duplicate tables, the computer program product including a computer readable storage medium having computer readable program code embodied therewith, the program instructions executable by a processor to cause the processor to:

convert syntactical data from relational tables from a schema of a database to knowledge graphs having semantic data;

map nodes in the knowledge graphs to sources in the relational tables;

apply graph matching to the knowledge graphs;

detect semantically equivalent tables from the relational tables by assessing a degree of matching between matched knowledge graphs based on equivalent data from common linked-pair nodes, defined from relationships, from the knowledge graphs; and

generate a corrective flag to eliminate the semantically equivalent tables from the relational tables from the schema of the database based on the degree of matching.

18. The computer program product of claim 17 , wherein the duplicate tables are present in a cloud based data lake.

19. The computer program product of claim 17 , wherein the relational tables comprise data arranged in columns.

20. The computer program product of claim 17 , wherein graph matching is a full match between at least two knowledge graphs, or a partial match between at least two knowledge graphs.

Assignments (1)

Continuity (1)

References Cited (36)