# US 2025/0094437 A1 · System and Method for Ingesting Data Based on Processed Metadata

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0094437 A1
- Document ID: US20250094437A1
- Status: Published application
- Kind: A1
- Application number: 18/469,906
- Filed: 2023-09-19
- Published: 2025-03-20
- Earliest claimed: 2023-09-19
- Assignees: The Toronto-Dominion Bank
- Applicants: The Toronto-Dominion Bank
- Inventors: Rajesh UPENDRAN
- Web page: https://patentgrep.com/patent/US20250094437A1

## CPC

- G06F 16/254: {Extract, transform and load [ETL] procedures, e.g. ETL data flows in data warehouses}
- G06F 16/285: {Clustering or classification}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A system, device and method are provided for assessing actions of authenticated persons within an enterprise system. The illustrative method includes extracting metadata comprising a plurality of categories from a plurality of data sources, and applying an unsupervised machine learning process to the extracted metadata. A plurality of clusters of the plurality of categories of the extracted metadata is generated, and thereafter one or more review criteria are applied thereto to generate curated clusters. The method includes training a supervised machine learning model with the curated clusters. The method includes, in response to receiving a new metadata input, processing the new metadata input with the trained supervised machine learning model. Data associated with the new metadata input is ingested based on respective clusters output by the trained supervised machine learning model for categories of the new metadata input.

## Claims

### Claim 1

A system for ingesting data based on processed metadata, the system comprising:
a processor;
a communications module coupled to the processor; and
a memory coupled to the processor, the memory storing computer executable instructions that when executed by the processor cause the processor to:
extract metadata comprising a plurality of categories from a plurality of data sources;
apply an unsupervised machine learning process to the extracted metadata to generate a plurality of clusters of the plurality of categories of the extracted metadata, the clusters defining attributes or parameters of metadata for the plurality of data sources as being related;
apply one or more review criteria to the generated plurality of clusters to generate a curated plurality of clusters;
train, with a supervised learning technique, a machine learning model with the curated plurality of clusters, the machine learning model being trained to predict a relevant cluster for categories of input metadata;
responsive to receiving a new metadata input, process the new metadata input with the trained supervised machine learning model; and
ingest data associated with the new metadata input based on respective clusters output by the trained supervised machine learning model for categories of the new metadata input.

### Claim 2 (depends on 1)

The system of claim 1, wherein the instructions cause the processor to:
extract the curated plurality of clusters into a relational database format.

### Claim 3 (depends on 2)

The system of claim 2, wherein the instructions cause the processor to:
receive a data structure indicating that metadata of a data source present in the relational database format is being altered; and
parse the relational database format to determine affected downstream applications.

### Claim 4 (depends on 1)

The system of claim 1, wherein applying one or more review criteria comprises the instructions causing the processor to:
generate an interface to receive input altering a composition of the plurality of clusters, the interface comprising one or more flagged entries for review.

### Claim 5 (depends on 1)

The system of claim 1, wherein the instructions cause the processor to:
receive data indicating errors associated with outputs of the trained supervised machine learning model;
retrain the trained supervised machine learning model based on the received data indicating errors; and
process new metadata with the re-trained supervised machine learning model.

### Claim 6 (depends on 1)

The system of claim 1, wherein the unsupervised machine learning process employs Kmodes.

### Claim 7 (depends on 1)

The system of claim 1, wherein the instructions cause the processor to:
automatically extract the new metadata input from a received data file from a data source to be ingested.

### Claim 8 (depends on 1)

The system of claim 1, wherein the extracted metadata provided to the unsupervised machine learning process comprises at least one of a data source, an attribute identifier, an associated application, an expected data type, and a data value.

### Claim 9 (depends on 1)

The system of claim 1, wherein the instructions cause the processor to:
provide a validator that is automated, the validator relying on relationships captured by the plurality of curated clusters to validate different data sources in a same cluster of the plurality of curated clusters.

### Claim 10 (depends on 1)

The system of claim 1, wherein to train the supervised machine learning model, the instructions cause the processor to:
train the supervised machine learning model with one or more pre-defined industry standards.

### Claim 11

A method for automating metadata processing, the method comprising:
extracting metadata from a plurality of data sources;
applying an unsupervised machine learning process to the extracted metadata to generate a plurality of clusters of the extracted metadata, the clusters defining attributes or parameters of metadata for the plurality of data sources as being related;
applying one or more review criteria to the generated plurality of clusters to generate a curated plurality of clusters;
training, with a supervised learning technique, a machine learning model with the curated plurality of clusters, the machine learning model being trained to predict a relevant cluster in response to input metadata;
responsive to receiving a new metadata input, processing the new metadata input with the trained supervised machine learning model; and
ingesting data associated with the new metadata input based on respective clusters output by the trained supervised machine learning model for categories of the new metadata input.

### Claim 12 (depends on 11)

The method of claim 11, comprising extracting the curated plurality of clusters into a relational database format.

### Claim 13 (depends on 12)

The method of claim 12, further comprising:
receiving a data structure indicating that metadata of a data source present in the relational database format is being altered; and
parsing the relational database format to determine affected downstream applications.

### Claim 14 (depends on 11)

The method of claim 11, wherein applying one or more review criteria comprises generating an interface to receive input altering a composition of the plurality of clusters, the interface comprising one or more flagged entries for review.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
receiving data indicating errors associated with outputs of the trained supervised machine learning model;
retraining the trained supervised machine learning model based on the received data indicating errors; and
processing new metadata with the re-trained supervised machine learning model.

### Claim 16 (depends on 11)

The method of claim 11, wherein the unsupervised machine learning process employs Kmodes.

### Claim 17 (depends on 11)

The method of claim 11, further comprising automatically extracting the new metadata input from a received data file from a data source to be ingested.

### Claim 18 (depends on 17)

The method of claim 17, wherein the extracted metadata provided to the unsupervised machine learning process comprises at least one of a data source, an attribute identifier, an associated application, an expected data type, and a data value range.

### Claim 19 (depends on 11)

The method of claim 11, further comprising:
providing a validator that is automated, the validator relying on relationships captured by the plurality of curated clusters to validate different data sources having categories grouped in a same cluster of the plurality of curated clusters.

### Claim 20

A non-transitory computer readable medium for automating metadata processing, the computer readable medium comprising computer executable instructions for:
extracting metadata from a plurality of data sources;
applying an unsupervised machine learning process to the extracted metadata to generate a plurality of clusters of the extracted metadata, the clusters defining attributes or parameters of metadata for the plurality of data sources as being related;
applying one or more review criteria to the generated plurality of clusters to generate a curated plurality of clusters;
training, with a supervised learning technique, a machine learning model with the curated plurality of clusters, the machine learning model being trained to predict a relevant cluster in response to input metadata;
responsive to receiving a new metadata input, processing the new metadata input with the trained supervised machine learning model; and
ingesting data associated with the new metadata input based on respective clusters output by the trained supervised machine learning model for categories of the new metadata input.
