# US 2025/0165713 A1 · Real-time named entity based transaction approval

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

- Number: US 2025/0165713 A1
- Document ID: US20250165713A1
- Status: Published application
- Kind: A1
- Application number: 19/033,267
- Filed: 2025-01-21
- Published: 2025-05-22
- Earliest claimed: 2021-06-17
- Assignees: Ramp Business Corporation
- Applicants: Ramp Business Corporation
- Inventors: Calvin Jun-Gong Lee; Beeral Dilip Patel; Paul Alexander Meier; Karim Atiyeh; Ariel Petren Langer; Ori Ephraim Goldfield; Geoffrey Jacques Charles
- Web page: https://patentgrep.com/patent/US20250165713A1

## CPC

- G06F 40/295: Named entity recognition
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06Q 40/02: Banking, e.g. interest calculation or account maintenance (credit or loans G06Q40/03)
- G06V 10/82: using neural networks
- G06V 30/414: Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text
- G06V 30/416: Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors
- G06F 40/205: Parsing
- G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
- G06F 40/279: Recognition of textual entities
- G06N 3/048: Activation functions
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A system for stable and streamlined transaction approval includes a rule management server and an authorizer server. The rule management server is configured to receive a transaction rule that includes a list of named entities specified by a client. The rule management server stores past transaction data and analyzes patterns in the past data to determine a named entity identification rule. The authorizer server receives a transaction data payload related to a pending transaction associated with the client and identifies a noisy data field in the transaction data payload. The noisy data field includes a representation of a particular named entity and one or more irrelevant strings. The authorizer server parses the particular named entity based on the named entity identification rule and determines whether the particular named entity is one of the named entities in the list. The authorizer server conducts a transaction evaluation using the result.

## Claims

### Claim 1

A computer-implemented method, comprising:
storing past transaction data associated with a plurality of clients;
generating training data from the past transaction data, the training data comprising labels of named entities associated with the past transaction data and noisy data fields in the past transaction data;
training a named-entity-identification model using the labels of the named entities and the noisy data fields in the training data;
receiving a transaction rule from a particular client;
receiving a transaction data payload related to a pending transaction associated with the particular client;
identifying a noisy data field in the transaction data payload, the noisy data field including a representation of a particular named entity and one or more irrelevant strings;
inputting the noisy data field into the named-entity-identification model to identify the particular named entity;
applying the transaction rule to the particular identified named entity; and
conducting a transaction evaluation to determine whether to approve the pending transaction based on the transaction rule.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein storing the past transaction data associated with the plurality of clients comprises:
retrieving transaction data payloads from a transaction terminal;
filtering the transaction data payloads to exclude irrelevant information; and
categorizing the transaction data payloads by client and merchant identifier.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein generating the training data from the past transaction data comprises:
identifying patterns in noisy data fields based on predefined rules;
assigning labels to named entities using historical data; and
validating the labeled data through a feedback loop.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, wherein training the named-entity-identification model comprises:
utilizing supervised learning techniques with labeled training samples;
iteratively optimizing the model using gradient descent; and
evaluating model accuracy based on cross-entropy loss.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein receiving the transaction rule from the particular client comprises:
presenting a graphical user interface for client input;
allowing the particular client to select from preconfigured rule templates; and
dynamically updating the rule based on client-specific requirements.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein identifying the noisy data field comprises:
applying heuristic algorithms to detect irrelevant substrings; and
standardizing a format of extracted data fields.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein inputting the noisy data field into the named-entity-identification model comprises:
tokenizing the noisy data field into substrings;
embedding substrings into a vector in a multi-dimensional vector space; and
normalizing the vector.

### Claim 8 (depends on 1)

The computer-implemented method of claim 1, wherein the named-entity-identification model utilizes a neural network architecture, the neural network architecture comprising:
convolutional layers for feature extraction;
recurrent layers for temporal pattern recognition; and
fully connected layers for classification.

### Claim 9 (depends on 1)

The computer-implemented method of claim 1, wherein applying the transaction rule comprises:
cross-referencing the identified named entity with a list stored in a database; and
triggering an alert responsive to the transaction violates a predefined threshold.

### Claim 10 (depends on 1)

The computer-implemented method of claim 1, wherein the named-entity-identification model is periodically updated to include:
new training data generated from cleared transactions; and
enhanced features derived from recent merchant identification patterns.

### Claim 11

A non-transitory computer-readable medium configured to store computer code comprising instructions, wherein the instructions, when executed by a processor, cause the processor to:
store past transaction data associated with a plurality of clients;
generate training data from the past transaction data, the training data comprising labels of named entities associated with the past transaction data and noisy data fields in the past transaction data;
train a named-entity-identification model using the labels of the named entities and the noisy data fields in the training data;
receive a transaction rule from a particular client;
receive a transaction data payload related to a pending transaction associated with the particular client;
identify a noisy data field in the transaction data payload, the noisy data field including a representation of a particular named entity and one or more irrelevant strings;
input the noisy data field into the named-entity-identification model to identify the particular named entity;
apply the transaction rule to the particular identified named entity; and
conduct a transaction evaluation to determine whether to approve the pending transaction based on the transaction rule.

### Claim 12 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to store the past transaction data associated with the plurality of clients comprises instructions to:
retrieve transaction data payloads from a transaction terminal;
filter the transaction data payloads to exclude irrelevant information; and
categorize the transaction data payloads by client and merchant identifier.

### Claim 13 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to generate the training data from the past transaction data comprises instructions to:
identify patterns in noisy data fields based on predefined rules;
assign labels to named entities using historical data; and
validate the labeled data through a feedback loop.

### Claim 14 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to train the named-entity-identification model comprises instructions to:
utilize supervised learning techniques with labeled training samples;
iteratively optimize the model using gradient descent; and
evaluate model accuracy based on cross-entropy loss.

### Claim 15 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to receive the transaction rule from the particular client comprises instructions to:
present a graphical user interface for client input;
allow the particular client to select from preconfigured rule templates; and
dynamically update the rule based on client-specific requirements.

### Claim 16 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to identify the noisy data field comprises instructions to:
apply heuristic algorithms to detect irrelevant substrings; and
standardize a format of extracted data fields.

### Claim 17 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to input the noisy data field into the named-entity-identification model comprises instructions to:
tokenize the noisy data field into substrings;
embed substrings into a vector in a multi-dimensional vector space; and
normalize the vector.

### Claim 18 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the named-entity-identification model utilizes a neural network architecture, the neural network architecture comprising:
convolutional layers for feature extraction;
recurrent layers for temporal pattern recognition; and
fully connected layers for classification.

### Claim 19 (depends on 11)

The non-transitory computer-readable medium of claim 11, wherein the instructions to apply the transaction rule comprises instructions to:
cross-reference the identified named entity with a list stored in a database; and
trigger an alert responsive to the transaction violates a predefined threshold.

### Claim 20

A system comprising:
a processor;
memory configured to store computer code comprising instructions, wherein the instructions, when executed by the processor, cause the processor to:
store past transaction data associated with a plurality of clients;
generate training data from the past transaction data, the training data comprising labels of named entities associated with the past transaction data and noisy data fields in the past transaction data;
train a named-entity-identification model using the labels of the named entities and the noisy data fields in the training data;
receive a transaction rule from a particular client;
receive a transaction data payload related to a pending transaction associated with the particular client;
identify a noisy data field in the transaction data payload, the noisy data field including a representation of a particular named entity and one or more irrelevant strings;
input the noisy data field into the named-entity-identification model to identify the particular named entity;
apply the transaction rule to the particular identified named entity; and
conduct a transaction evaluation to determine whether to approve the pending transaction based on the transaction rule.