# US 2026/0301594 A1 · Method and server device for interactive learning

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

- Number: US 2026/0301594 A1
- Document ID: US20260301594A1
- Status: Published application
- Kind: A1
- Application number: 19/359,919
- Filed: 2025-10-16
- Published: 2026-10-01
- Earliest claimed: 2025-03-27
- Assignees: National Taiwan Normal University
- Applicants: National Taiwan Normal University
- Inventors: Jon-Chao HONG; Huan YEH; Chi Chin WONG
- Web page: https://patentgrep.com/patent/US20260301594A1

## CPC

- G09B 7/02: of the type wherein the student is expected to construct an answer to the question which is presented or wherein the machine gives an answer to the question presented by a student
- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G09B: EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS (devices for psychotechnics or for testing reaction times A61B5/16; games, sports, amusements A63; projectors, projector screens G03B)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method for providing interactive learning is implemented using a server device that stores a large language model (LLM) and entries of knowledge data that correspond with a plurality of knowledge texts, respectively, the method includes: in response to receipt of a question string, obtaining an entry of target knowledge data and an answer string; obtaining and transmitting an inquiring string that is different from an answer of a preset question, and transmitting the inquiring string; in response to receipt of a first reply string, determining whether the first reply string corresponds with the answer of the preset question; in the case that the first reply string does not correspond with the answer of the preset question, obtaining a hinting string based on the first reply string and the entry of target knowledge data, the hinting string including hints to the answer.

## Claims

### Claim 1

A method for interactive learning for a user to obtain a correct answer to a preset question, the method being implemented using a server device that is in communication with a user device held by the user, the server device storing a large language model (LLM) and a plurality of entries of knowledge data that are related to a plurality of knowledge texts, respectively, the method comprising steps of:
a) in response to receipt of a question string from the user device, obtaining at least one entry of target knowledge data that is selected from the plurality of entries of knowledge data and that is associated with the correct answer of the present question;
b) using the LLM to obtain an answer string from the entry of target knowledge data;
c) using the LLM to obtain an inquiring string based on the entry of target knowledge data, the inquiring string being different from the correct answer of the preset question, and transmitting the inquiring string to the user device;
d) in response to receipt of a first reply string from the user device, determining whether the first reply string conforms with the correct answer of the preset question;
e) in the case that the first reply string does not conform with the correct answer of the preset question, using the LLM to obtain a hinting string based on the first reply string and the entry of target knowledge data, and transmitting the hinting string to the user device, the hinting string including hints to instruct the user to continue thinking for the correct answer to the preset question;
f) in response to receipt of a second reply string from the user device, determining whether the second reply string conforms with the correct answer of the preset question; and
g) in the case it is determined that the second reply string does not conform with the correct answer of the preset question, obtaining and transmitting one of a first partial answer string and the answer string to the user device, the first partial answer string being obtained based on one of the feedback strings and the answer string.

### Claim 2 (depends on 1)

The method as claimed in claim 1, further comprising:
prior to step a), generating, for each of the plurality of entries of knowledge data, a corresponding knowledge graph;
i) after step a), determining whether the entry of target knowledge data has relevance based on the knowledge graph that corresponds with the entry of target knowledge data;
j) in the case that it is determined that the entry of target knowledge data has relevance, assigning the entry of target knowledge data into a cluster, based on the knowledge graph; and
k) using the LLM to generate a partitioned string that is associated with the entry of target knowledge data contained in the cluster, and transmitting the partitioned string to the user device.

### Claim 3 (depends on 2)

The method as claimed in claim 2, wherein step i) includes identifying a plurality of entities from the knowledge graph that corresponds with the entry of target knowledge data, and determining whether there exists a relation between any two among the plurality of entities,
wherein, in the case there exists a relation between any two among the plurality of entities, it is determined that the entry of target knowledge data has relevance.

### Claim 4 (depends on 2)

The method as claimed in claim 2, further comprising, after step i):
l) in the case that it is determined that the entry of target knowledge data does not have relevance, using the LLM to generate an auxiliary string that is associated with the knowledge graph corresponding with the entry of target knowledge data, and that is for assisting the user to better remember the content of the knowledge text; and
m) transmitting the auxiliary string to the user device.

### Claim 5 (depends on 4)

The method as claimed in claim 4, wherein step l) includes:
determining whether to use a first manner or a second manner to generate the auxiliary string;
in the first manner, using the LLM to generate the auxiliary string that incorporates the content of the entry of target knowledge data in a fictional story; and
in the second manner, using the LLM to generate the auxiliary string that adjusts the content of the entry of target knowledge data using a mnemonic device.

### Claim 6 (depends on 1)

The method as claimed in claim 1, the server device further storing a plurality of positive terms for encouraging the user, a plurality of feedback categories each being associated with a specific feedback style, each of the feedback categories including a plurality of feedback strings, the method further comprising, prior to step c):
n) processing at least the question string to obtain a first result associated with an emotional state of the user using an emotion detection algorithm, the first result indicating whether the user is in a negative emotional state;
wherein step c) includes sub-steps of
c-1) in the case that the user is not in the negative emotional state, obtaining the inquiring string based on one of the feedback strings and the entry of target knowledge data, the one of the feedback strings being selected from a current feedback category which is one of the plurality of feedback categories;
c-2) in the case that the user is in the negative emotional state, obtaining the inquiring string based on one of the feedback strings, the entry of target knowledge data, and one of the positive terms, the one of the feedback strings being selected from an alternative feedback category, which is one of the plurality of feedback categories and is different from the current feedback category; and
c-3) setting the alternative feedback category as the current feedback category.

### Claim 7 (depends on 6)

The method as claimed in claim 6, the user device including a user-end camera module that periodically captures facial images of the user and being configured to transmit the facial images to the server device, wherein step n) includes sub-steps of:
n-1) using the emotion detection algorithm to process the question string to obtain a first text result associated with an emotional state of the user;
n-2) selecting a number of to-be-processed images that are the facial images captured by the user-end camera module during a period of time from a time point prior to receiving the question string to a time point when the server device receives the question string;
n-3) processing the number of to-be-processed images to obtain a first image result associated with the emotional state of the user; and
n-4) using the first text result and the first image result to obtain the first result.

### Claim 8 (depends on 1)

The method as claimed in claim 1, the server device further storing a plurality of positive terms for encouraging the user, a plurality of feedback categories each being associated with a specific feedback style, each of the feedback categories including a plurality of feedback strings, the method further comprising, prior to step e):
o) processing at least the first reply string to obtain a second result associated with an emotional state of the user using an emotion detection algorithm, the second result indicating whether the user is in a negative emotional state;
wherein step e) includes the sub-steps of
e-1) in the case that the user is not in the negative emotional state, obtaining the hinting string based on one of the feedback strings, the first reply string, and the entry of target knowledge data, the one of the feedback strings being selected from a current feedback category which is one of the plurality of feedback categories;
e-2) in the case that the user is in the negative emotional state, obtaining the hinting string based on one of the feedback strings, the first reply string, the entry of target knowledge data and one of the positive terms, the one of the feedback strings being selected from an alternative feedback category, which is one of the plurality of feedback categories and is different from the current feedback category; and
e-3) setting the alternative feedback category as the current feedback category.

### Claim 9 (depends on 8)

The method as claimed in claim 8, the user device including a user-end camera module that periodically captures facial images of the user and that is configured to transmit the facial images to the server device, wherein step o) includes the sub-steps of:
o-1) using the emotion detection algorithm to process the question string to obtain a second text result associated with an emotional state of the user;
o-2) selecting a number of to-be-processed images that are the facial images captured by the user-end camera module during a period of time from a time point prior to receiving the first reply string to a time point when the server device receives the first reply string;
o-3) processing the number of to-be-processed images to obtain a second image result associated with the emotional state of the user; and
o-4) using the second text result and the second image result to obtain the second result.

### Claim 10 (depends on 1)

The method as claimed in claim 1, the server device further storing a learning record associated with the user, a plurality of suggestion strings that correspond with a plurality of different learning styles, respectively, and a style classification algorithm that is configured to classify the user into one of the plurality of different learning styles, the method further comprising:
in response to receipt of a filled-out learning questionnaire from the user device, executing the style classification algorithm so as to classify the user into one of the plurality of different learning styles based on the filled-out learning questionnaire and the learning record associated with the user; and
generating a learning suggestion for the user based on one of the suggestion strings that corresponds with the one of the plurality of different learning styles associated with the user, and transmitting the learning suggestion to the user device.

### Claim 11

A server device for interactive learning for a user to obtain a correct answer to a preset question, the server device comprising a server communication module for establishing a communication with a user device held by the user, a server data storage module that stores a large language model (LLM) and a plurality of entries of knowledge data that are related to a plurality of knowledge texts, respectively, and a server processor that is electrically connected to the server communication module and the server data storage module, and that is programmed to:
in response to receipt of a question string from the user device, obtain at least one entry of target knowledge data that is selected from the plurality of entries of knowledge data and that is associated with the correct answer of the present question;
use the LLM to obtain an answer string from the entry of target knowledge data;
use the LLM to obtain an inquiring string based on the entry of target knowledge data, the inquiring string being different from the correct answer of the preset question, and control the server communication module to transmit the inquiring string to the user device;
in response to receipt of a first reply string from the user device, determine whether the first reply string conforms with the correct answer of the preset question;
in the case that the first reply string does not conform with the correct answer of the preset question, use the LLM to obtain a hinting string based on the first reply string and the entry of target knowledge data, and control the server communication module to transmit the hinting string to the user device, the hinting string including hints to instruct the user to continue thinking for the correct answer to the preset question;
in response to receipt of a second reply string from the user device, determine whether the second reply string conforms with the correct answer of the preset question; and
in the case it is determined that the second reply string does not conform with the correct answer of the preset question, control the server communication module to transmit one of a first partial answer string and the answer string to the user device, the first partial answer string being obtained based on one of the feedback strings and the answer string.

### Claim 12 (depends on 11)

The server device as claimed in claim 11, wherein the server processor is further programmed to:
prior to obtaining the at least one entry of target knowledge data, generate, for each of the plurality of entries of knowledge data, a corresponding knowledge graph;
after obtaining the at least one entry of target knowledge data, determine whether the entry of target knowledge data has relevance based on the knowledge graph that corresponds with the entry of target knowledge data;
in the case that it is determined that the entry of target knowledge data has relevance, assign the entry of target knowledge data into a cluster, based on the knowledge graph; and
use the LLM to generate a partitioned string that is associated with the entry of target knowledge data contained in the cluster, and transmitting the partitioned string to the user device.

### Claim 13 (depends on 12)

The server device as claimed in claim 12, wherein the server processor determines whether the entry of target knowledge data has relevance by identifying a plurality of entities from the knowledge graph that corresponds with the entry of target knowledge data, and determining whether there exists a relation between any two among the plurality of entities,
wherein, in the case there exists a relation between any two among the plurality of entities, it is determined that the entry of target knowledge data has relevance.

### Claim 14 (depends on 12)

The server device as claimed in claim 12, wherein the server processor is further programmed to, after determining whether the entry of target knowledge data has relevance:
in the case that it is determined that the entry of target knowledge data does not have relevance, use the LLM to generate an auxiliary string that is associated with the knowledge graph corresponding with the entry of target knowledge data, and that is for assisting the user to better remember the content of the knowledge text; and
control the server communication module to transmit the auxiliary string to the user device.

### Claim 15 (depends on 14)

The server device as claimed in claim 14, wherein the server processor generates the auxiliary string by:
determining whether to use a first manner or a second manner to generate the auxiliary string;
in the first manner, using the LLM to generate the auxiliary string that incorporates the content of the entry of target knowledge data in a fictional story; and
in the second manner, using the LLM to generate the auxiliary string that adjusts the content of the entry of target knowledge data using a mnemonic device.

### Claim 16 (depends on 11)

The server device as claimed in claim 11, wherein the server data storage module further stores a plurality of positive terms for encouraging the user, a plurality of feedback categories each being associated with a specific feedback style, each of the feedback categories including a plurality of feedback strings, the server processor is further programmed to, prior to obtain the inquiring string:
process at least the question string to obtain a first result associated with an emotional state of the user using an emotion detection algorithm, the first result indicating whether the user is in a negative emotional state;
wherein the server processor obtains the inquiring string by
in the case that the user is not in the negative emotional state, obtaining the inquiring string based on one of the feedback strings and the entry of target knowledge data, the one of the feedback strings being selected from a current feedback category which is one of the plurality of feedback categories;
in the case that the user is in the negative emotional state, obtaining the inquiring string based on one of the feedback strings, the entry of target knowledge data, and one of the positive terms, the one of the feedback strings being selected from an alternative feedback category, which is one of the plurality of feedback categories and is different from the current feedback category; and
setting the alternative feedback category as the current feedback category.

### Claim 17 (depends on 16)

The server device as claimed in claim 16, the user device including a user-end camera module that periodically captures facial images of the user and being configured to transmit the facial images to the server device, wherein the server processor obtains a first result by:
using the emotion detection algorithm to process the question string to obtain a first text result associated with an emotional state of the user;
selecting a number of to-be-processed images that are the facial images captured by the user-end camera module during a period of time from a time point prior to receiving the question string to a time point when the server device receives the question string;
processing the number of to-be-processed images to obtain a first image result associated with the emotional state of the user; and
using the first text result and the first image result to obtain the first result.

### Claim 18 (depends on 11)

The server device as claimed in claim 11, wherein the server data storage module further stores a plurality of positive terms for encouraging the user, a plurality of feedback categories each being associated with a specific feedback style, each of the feedback categories including a plurality of feedback strings, the server processor is further programmed to, prior to obtaining the hinting string:
process at least the first reply string to obtain a second result associated with an emotional state of the user using an emotion detection algorithm, the second result indicating whether the user is in a negative emotional state;
wherein the server processor obtains the hinting string by
in the case that the user is not in the negative emotional state, obtaining the hinting string based on one of the feedback strings, the first reply string, and the entry of target knowledge data, the one of the feedback strings being selected from a current feedback category which is one of the plurality of feedback categories;
in the case that the user is in the negative emotional state, obtaining the hinting string based on one of the feedback strings, the first reply string, the entry of target knowledge data and one of the positive terms, the one of the feedback strings being selected from an alternative feedback category, which is one of the plurality of feedback categories and is different from the current feedback category; and
setting the alternative feedback category as the current feedback category.

### Claim 19 (depends on 18)

The server device as claimed in claim 18, the user device including a user-end camera module that periodically captures facial images of the user and that is configured to transmit the facial images to the server device, wherein the server processor obtains the second result by:
using the emotion detection algorithm to process the question string to obtain a second text result associated with an emotional state of the user;
selecting a number of to-be-processed images that are the facial images captured by the user-end camera module during a period of time from a time point prior to receiving the first reply string to a time point when the server device receives the first reply string;
processing the number of to-be-processed images to obtain a second image result associated with the emotional state of the user; and
using the second text result and the second image result to obtain the second result.

### Claim 20 (depends on 11)

The server device as claimed in claim 11, wherein the server data storage module further stores a learning record associated with the user, a plurality of suggestion strings that correspond with a plurality of different learning styles, respectively, and a style classification algorithm that is configured to classify the user into one of the plurality of different learning styles, the server processor is further programmed to:
in response to receipt of a filled-out learning questionnaire from the user device, execute the style classification algorithm so as to classify the user into one of the plurality of different learning styles based on the filled-out learning questionnaire and the learning record associated with the user; and
generate a learning suggestion for the user based on one of the suggestion strings that corresponds with the one of the plurality of different learning styles associated with the user, and control the server communication module to transmit the learning suggestion to the user device.
