# US 2025/0094633 A1 · Data processing method and system for interacting with large language models

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

- Number: US 2025/0094633 A1
- Document ID: US20250094633A1
- Status: Published application
- Kind: A1
- Application number: 18/762,244
- Filed: 2024-07-02
- Published: 2025-03-20
- Earliest claimed: 2023-09-18
- Assignees: Droidtown Linguistic Technology Co., Ltd.
- Applicants: Droidtown Linguistic Technology Co., Ltd.
- Inventors: WEN-JET WANG
- Web page: https://patentgrep.com/patent/US20250094633A1

## CPC

- G06F 21/6254: {by anonymising data, e.g. decorrelating personal data from the owner's identification}
- G06F 40/30: Semantic analysis
- G06F 40/35: Discourse or dialogue representation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A data processing method for interacting with large language models, comprising obtaining an original text, analyzing the original text and determining confidential data items in the original text, generating a semantic model representing the relations across the items, converting the confidential data items in the original text into a common word to generate a secure input text, inputting the secure input text to a large language model for processing, receiving a reply text generated by the large language model based on the secure input text, and converting the reply text into a target reply text corresponding to the original text according to the common word and the semantic model.

## Claims

### Claim 1

A data processing method for interacting with large language models, comprising:
obtaining an original text;
analyzing the original text and determining a confidential data item in the original text;
generating a semantic model according to sentence structure, semantic content and contextual information associated with the confidential data item in the original text and converting the confidential data item in the original text into a common word for masking, so as to generate a secure input text;
inputting the secure input text to a large language model for processing;
receiving a reply text generated by the large language model based on the secure input text; and
converting the reply text into a target reply text corresponding to the original text for unmasking according to the common word and the semantic model.

### Claim 2 (depends on 1)

The data processing method of claim 1, further comprising:
performing a de-identifying and masking process on the original text by converting the confidential data item into the common word according to at least one of word class, word property, sentence element category and contextual information of the confidential data item, and accordingly generating the secure input text.

### Claim 3 (depends on 2)

The data processing method of claim 2, further comprising:
converting the common word into the confidential data item according to at least one of word class, word property, sentence element category and contextual information of the common word for generating the target reply text.

### Claim 4 (depends on 1)

The data processing method of claim 1, wherein the confidential data item is different from the common word.

### Claim 5 (depends on 1)

The data processing method of claim 1, wherein the confidential data item belongs to the same word class as the common word.

### Claim 6 (depends on 1)

The data processing method of claim 1, wherein the target reply text conforms to a natural language structure.

### Claim 7 (depends on 1)

The data processing method of claim 1, wherein the large language model is ChatGPT language model, Gemini (formerly known as Bard) language model, Poe language model or Claude language model.

### Claim 8

A data processing system for interacting with large language models, comprising:
an information system unit, configured to store instructions;
a calculation core module, configured to execute the instructions of:
obtaining an original text;
analyzing the original text and determining a confidential data item in the original text;
generating a semantic model according to sentence structure, semantic content and contextual information associated with the confidential data item in the original text and converting the confidential data item in the original text into a common word for masking, so as to generate a secure input text;
inputting the secure input text to a large language model for processing;
receiving a reply text generated by the large language model based on the secure input text; and
converting the reply text into a target reply text corresponding to the original text for unmasking according to the common word and the semantic model.

### Claim 9 (depends on 8)

The data processing system of claim 8, wherein the instructions further comprise:
performing a de-identifying and masking process on the original text by converting the confidential data item into the common word according to at least one of word class, word property, sentence element category and contextual information of the confidential data item, and accordingly generating the secure input text.

### Claim 10 (depends on 9)

The data processing system of claim 9, wherein the instructions further comprise:
converting the common word into the confidential data item according to at least one of word class, word property, sentence element category and contextual information of the common word for generating the target reply text.

### Claim 11 (depends on 8)

The data processing system of claim 8, wherein the confidential data item is different from the common word.

### Claim 12 (depends on 8)

The data processing system of claim 8, wherein the confidential data item belongs to the same word class as the common word.

### Claim 13 (depends on 8)

The data processing system of claim 8, wherein the target reply text conforms to a natural language structure.