Granted Patent B2

US 12,625,872 · App. 18/659,616 · Granted May 12, 2026

Inventors: Matthew J. Glickman (Larchmont, NY); Orestis Kostakis (Redmond, WA); Justin Langseth (Kailua, HI)

Assignee: Snowflake Inc.

G06F16/24568G06F16/244G06F16/2456G06F16/24564

Abstract

Claims (54)

1 . A method comprising:

monitoring, by at least one hardware processor, a first user interaction with a first dataset on a database system, the first dataset being associated with a first user;

receiving, from the first user, feedback regarding a quality of a results dataset generated from an executed overlap query, the results dataset comprising a joined dataset of data from the first dataset of the first user and data from a second dataset of a second user;

based on the feedback, generating a similarity score dataset indicating a similarity between the first dataset and the second dataset;

generating, using the similarity score dataset, a new set of overlap queries specifying join operations between the first dataset and the second dataset;

executing the new set of overlap queries to materialize a new joined dataset comprising the data from the first dataset and the data from the second dataset, the materializing comprises persisting the new joined dataset as a physical table indexed and stored in the database system for subsequent reuse;

generating a new results dataset by applying the new set of overlap queries to the new joined dataset; and

providing, to the first user, the new results dataset comprising a refinement recommendation for the overlap queries based on additional user feedback, the refinement recommendation comprising a ranked set of proposed modifications to at least one join operation associated with the overlap queries.

2 . The method of claim 1 , wherein each overlap query is generated based on non-matching data between the first dataset and a corresponding dataset of the new joined dataset, the non-matching data comprises non-matching columns between the first dataset and the new joined dataset.

3 . The method of claim 2 , wherein generating the new set of overlap queries comprises identifying, from an overlap query repository, an overlap query that is associated with the non-matching data.

4 . The method of claim 1 , further comprising:

prior to generating the new joined dataset, executing a data cleaning process on the first dataset and the second dataset to ensure data quality and consistency.

5 . The method of claim 1 , wherein the feedback from the first user is received via an interactive user interface and further comprises:

enabling the first user to rate the quality of the results dataset on a predefined scale.

6 . The method of claim 1 , wherein the new set of overlap queries is generated based on a machine learning model that predicts potential utility of overlap datasets based on historical user feedback and query outcomes.

7 . The method of claim 6 , wherein the machine learning model is retrained periodically with new user feedback and results datasets to improve accuracy of the new set of overlap queries.

8 . The method of claim 1 , further comprising:

categorizing each dataset in the database system by a semantic type of a data field within each dataset, wherein the semantic type includes at least one of a ZIP code, a date, or a numerical identifier.

9 . The method of claim 8 , wherein generating the similarity score dataset includes comparing the semantic type of the first dataset with the semantic type of the second dataset to identify matching and non-matching data fields.

10 . A system comprising:

one or more hardware processors; and

one or more computer-readable mediums storing instructions that, when executed by the one or more hardware processors, cause the system to perform operations comprising:

monitoring a first user interaction with a first dataset on a database system, the first dataset being associated with a first user;

receiving, from the first user, feedback regarding a quality of a results dataset generated from an executed overlap query, the results dataset comprising a joined dataset of data from the first dataset of the first user and data from a second dataset of a second user;

based on the feedback, generating a similarity score dataset indicating a similarity between the first dataset and the second dataset;

generating, using the similarity score dataset, a new set of overlap queries specifying join operations between the first dataset and the second dataset;

executing the new set of overlap queries to materialize a new joined dataset comprising the data from the first dataset and the data from the second dataset, the materializing comprises persisting the new joined dataset as a physical table indexed and stored in the database system for subsequent reuse;

generating a new results dataset by applying the new set of overlap queries to the new joined dataset; and

providing, to the first user, the new results dataset comprising a refinement recommendation for the overlap queries based on additional user feedback, the refinement recommendation comprising a ranked set of proposed modifications to at least one join operation associated with the overlap queries.

11 . The system of claim 10 , wherein each overlap query is generated based on non- matching data between the first dataset and a corresponding dataset of the new joined dataset, the non-matching data comprises non-matching columns between the first dataset and the new joined dataset.

12 . The system of claim 10 , the operations further comprising:

prior to generating the new joined dataset, executing a data cleaning process on the first dataset and the second dataset to ensure data quality and consistency.

13 . The system of claim 10 , wherein the feedback from the first user is received via an interactive user interface and the operations further comprising:

enabling the first user to rate the quality of the results dataset on a predefined scale.

14 . The system of claim 10 , wherein the new set of overlap queries is generated based on a machine learning model that predicts potential utility of overlap datasets based on historical user feedback and query outcomes, the machine learning model is retrained periodically with new user feedback and results datasets to improve accuracy of the new set of overlap queries.

15 . The system of claim 10 , the operations further comprising:

categorizing each dataset in the database system by a semantic type of a data field within each dataset, wherein the semantic type includes at least one of a ZIP code, a date, or a numerical identifier; and

comparing the semantic type of the first dataset with the semantic type of the second dataset to identify matching and non-matching data fields.

16 . A machine-readable storage device embodying instructions that, when executed by a machine, cause the machine to perform operations comprising:

monitoring a first user interaction with a first dataset on a database system, the first dataset being associated with a first user;

receiving, from the first user, feedback regarding a quality of a results dataset generated from an executed overlap query, the results dataset comprising a joined dataset of data from the first dataset of the first user and data from a second dataset of a second user;

based on the feedback, generating a similarity score dataset indicating a similarity between the first dataset and the second dataset;

generating, using the similarity score dataset, a new set of overlap queries specifying join operations between the first dataset and the second dataset;

executing the new set of overlap queries to materialize a new joined dataset comprising the data from the first dataset and the data from the second dataset, the materializing comprises persisting the new joined dataset as a physical table indexed and stored in the database system for subsequent reuse;

generating a new results dataset by applying the new set of overlap queries to the new joined dataset; and

providing, to the first user, the new results dataset comprising a refinement recommendation for the overlap queries based on additional user feedback, the refinement recommendation comprising a ranked set of proposed modifications to at least one join operation associated with the overlap queries.

17 . The machine-readable storage device of claim 16 , wherein each overlap query is generated based on non-matching data between the first dataset and a corresponding dataset of the new joined dataset, the non-matching data comprises non-matching columns between the first dataset and the new joined dataset.

18 . The machine-readable storage device of claim 16 , the operations further comprising:

prior to generating the new joined dataset, executing a data cleaning process on the first dataset and the second dataset to ensure data quality and consistency.

19 . The machine-readable storage device of claim 16 , wherein the feedback from the first user is received via an interactive user interface and the operations further comprising:

enabling the first user to rate the quality of the results dataset on a predefined scale.

20 . The machine-readable storage device of claim 16 , the operations further comprising:

categorizing each dataset in the database system by a semantic type of a data field within each dataset, wherein the semantic type includes at least one of a ZIP code, a date, or a numerical identifier; and

comparing the semantic type of the first dataset with the semantic type of the second dataset to identify matching and non-matching data fields.

Assignments (1)

Continuity (2)

References Cited (45)