	US 12,373,905 B2
	Method of recommending preceding educational content and educational content recommendation device for recommending preceding educational content
Hyun Bin Loh, Seoul (KR)
Assigned to RIIID INC., Seoul (KR)
Filed by RIIID INC., Seoul (KR)
Filed on Nov. 8, 2022, as Appl. No. 17/982,648.
Claims priority of application No. 10-2021-0152036 (KR), filed on Nov. 8, 2021.
Prior Publication US 2023/0144716 A1, May 11, 2023
Int. Cl. G06Q 50/20 (2012.01); G06F 40/40 (2020.01)

CPC G06Q 50/205 (2013.01) [G06F 40/40 (2020.01)]

8 Claims

1. A method of recommending preceding educational content to be learned in advance by a user to solve educational content by an educational content recommendation device for acquiring the educational content and recommending the preceding educational content, wherein the educational content recommendation device comprises a memory storing computer-readable instructions and a controller configured to execute the instructions to perform the method comprising:

acquiring a language model of which training has been completed;

updating the language model by tuning the language model to acquire a target language model; and

determining the preceding educational content through the target language model,

wherein the acquirement of the target language model comprises:

acquiring an educational data set including first clustering data and second clustering data; and

updating the language model to predict a probability that the first clustering data is a next token of the second clustering data-,

wherein the determining of the preceding educational content comprises:

acquiring target educational content,

acquiring an educational content database including one or more pieces of candidate educational content,

calculating probabilities that the target educational content is a next token of the pieces of candidate educational content included in the educational content database through the target language model, and

determining the preceding educational content among the one or more pieces of candidate educational content included in the educational content database on the basis of the calculated probabilities,

wherein the educational content database includes first candidate educational content and second candidate educational content, and

wherein the calculating of the probabilities that the target educational content is the next token comprises:

calculating a first probability that the target educational content is a next token of the first candidate educational content, and

calculating a second probability that the target educational content is a next token of the second candidate educational content,

wherein the determining of the preceding educational content further comprises:

comparing the first probability with the second probability, and

selecting the candidate educational content having a larger probability value as the preceding educational content according to a result of the comparison, and

wherein the tuning of the language model comprises tuning the language model by using a next-token prediction algorithm which is a natural language processing technique.