Identifying and analyzing actions from vector representations of alphanumeric characters using a large language model

Patent metadata

Field	Value
Publication number	US-12505291-B2
Application number	US-202519019365-A
Country	US
Kind code	B2
Filing date	Jan 13, 2025
Priority date	Apr 11, 2024
Publication date	Dec 23, 2025
Grant date	Dec 23, 2025

Abstract

The systems and methods disclosed herein receive an output generation request from that includes input for generating an output using a language model. The input includes a set of alphanumeric characters associated with operative standards for a first set of actions. The system divides the set of alphanumeric characters into text subsets. For each text subset, a vector representation is determined. Prompts are created for each vector representation including the set of alphanumeric characters, query contexts, keywords, and/or the text subset. Each vector representation's prompt is input into the language model, which generates a second set of actions of related actions, where subsequently generated actions are based on prior generated actions. The system aggregates the second set of actions into a third set of actions and displays a graphical layout. The graphical layout displays a representation of the set of alphanumeric characters and the corresponding actions.

First claim

We claim:

1 . A non-transitory computer-readable storage medium comprising instructions stored thereon, wherein the instructions when executed by at least one data processor of a system, cause the system to:
obtain (1) an output generation request for generation of an output and (2) one or more indicators of a first set of actions,
wherein the output generation request includes one or more indicators of a set of guidelines, and
wherein the first set of actions are configured to adhere to constraints of the set of guidelines;
using the one or more indicators of the set of guidelines, determine a set of inputs for one or more guidelines in the set of guidelines,
wherein at least one input of the set of inputs for a particular guideline includes one or more of: (1) the particular guideline, (2) a set of pre-loaded query context defining the first set of actions, (3) a set of keywords associated with a set of textual content of corresponding guidelines of the first set of actions, or (4) the set of textual content; and
associate, using an AI model, the one or more guidelines of the set of guidelines to one or more actions in the first set of actions by:
supplying one or more inputs of the set of inputs of a corresponding guideline into the AI model, and
receiving, from the AI model, a generated second set of actions including the one or more actions of the first set of actions associated with the corresponding guideline of the set of guidelines.

2 . The non-transitory, computer-readable storage medium of claim 1 , wherein the set of guidelines includes one or more of: text, image, audio, or video format data.

3 . The non-transitory, computer-readable storage medium of claim 1 , wherein the AI model is a first AI model, and wherein the instructions further cause the system to:
supply the set of guidelines into a second AI model; receive, from the second AI model, a set of summaries summarizing the set of guidelines; and include at least one summary in the set of inputs.

4 . The non-transitory, computer-readable storage medium of claim 1 , wherein the instructions further cause the system to:
identify sensitive information within the set of guidelines using at least one predefined indicator of the sensitive information; and mask the identified sensitive information.

5 . The non-transitory, computer-readable storage medium of claim 1 , wherein the instructions further cause the system to:
receive an indicator of a type of operation associated with the set of guidelines; and obtain, via an application programming interface (API), the set of guidelines.

6 . The non-transitory, computer-readable storage medium of claim 1 , wherein the instructions further cause the system to:
associate each guideline with a type of guideline; and generate the second set of actions using the type of guideline of each guideline in the set of guidelines.

7 . The non-transitory, computer-readable storage medium of claim 1 , wherein the instructions further cause the system to:
validate the generated second set of actions by:
comparing each action in the second set of actions against predefined validation criteria; and
generating a validation score for each action based on the comparisons.

8 . A system comprising:
at least one hardware processor; and at least one non-transitory memory storing instructions, which, when executed by the at least one hardware processor, cause the system to:
obtain (1) an output generation request for generation of an output and (2) one or more indicators of a first set of actions,
wherein the output generation request includes one or more indicators of a set of guidelines, and
wherein the first set of actions are configured to adhere to constraints of the set of guidelines;

using the one or more indicators of the set of guidelines, determine a set of information for one or more guidelines in the set of guidelines,
wherein at least a portion of the set of information for a particular guideline includes one or more of: (1) the particular guideline, (2) a set of pre-loaded query context defining the first set of actions, (3) a set of keywords associated with a set of textual content of corresponding guidelines of the first set of actions, or (4) the set of textual content; and

associate, using an AI model, the one or more guidelines of the set of guidelines to one or more actions in the first set of actions by:
supplying the set of information of a corresponding guideline into the AI model, and
receiving, from the AI model, a generated second set of actions including the one or more actions of the first set of actions associated with the corresponding guideline of the set of guidelines.

9 . The system of claim 8 , wherein the system is further caused to:
calculate a set of token counts within the set of guidelines; and partition the set of guidelines into a set of segments based on predetermined character limits of the set of token counts.

10 . The system of claim 8 , wherein the one or more indicators of the set of guidelines include one or more of:
electronic tags associated with guideline types, metadata indicating operative standard categories, or scenario attributes defining one or more operational contexts.

11 . The system of claim 8 , wherein the set of information include one or more of:
vector representations of alphanumeric characters, textual content extracted from the guidelines using natural language processing, or pre-established mappings between operative standards and identified gaps.

12 . The system of claim 8 , wherein the system is further caused to:
generate a confidence score for each association between the one or more guidelines and a third set of actions; and generate the second set of actions by filtering the associations based on a predetermined confidence threshold.

13 . The system of claim 8 , wherein the system is further caused to:
detect a set of changes in the set of guidelines; and automatically update the second set of actions based on the detected changes.

14 . The system of claim 8 , wherein the system is further caused to:
generate a set of metadata documenting the association between the one or more guidelines and the one or more actions,
wherein the set of metadata includes one or more of: a timestamp, model version, or a confidence metric of the association.

15 . A method performed by a computer device for associating guidelines with one or more actions, the method comprising:
obtain (1) an output generation request for generation of an output and (2) one or more indicators of a first set of actions,
wherein the output generation request includes one or more indicators of a set of guidelines, and
wherein the first set of actions are configured to adhere to constraints of the set of guidelines;
using the one or more indicators of the set of guidelines, determine a set of information for one or more guidelines in the set of guidelines,
wherein the set of information for a particular guideline includes one or more of: (1) the particular guideline, (2) a set of pre-loaded query context defining the first set of actions, (3) a set of keywords associated with a set of textual content of corresponding guidelines of the first set of actions, or (4) the set of textual content; and
causing an AI model to associate the one or more guidelines of the set of guidelines to one or more actions in the first set of actions using at least a portion of the set of information of a corresponding guideline.

16 . The method of claim 15 , further comprising:
gene

Assignees

• Citibank Na

Inventors

• Malviya Shardul
• Liao Wayne
• Jain Deepak
• Cory Samantha
• Saternus Mariusz
• Lewandowski Daniel
• Rath Biraj Krushna
• Murray Stuart
• Davies Philip
• Jain Payal
• Maonah Tariq Husayn

Classifications

• G06F40/216 Primary

  using statistical methods · CPC title