KR20240006913A - Device for Evaluating Status of Student By Way Of Multiple Artificial Neural Networks - Google Patents

Abstract

One purpose of the present invention is to enable the learner's current learning state to be predicted based on the learner's learning record.
To this end, a learner status diagnosis device including a plurality of artificial neural network modules according to an embodiment of the present invention, the plurality of artificial neural network modules include a transformer-based artificial neural network module, and the transformer-based artificial neural network module. It includes a plurality of LSTM (Long Short-Term Memory) modules installed at the front and rear ends of the , wherein the plurality of artificial neural network modules receive learning records from a learner client and process the input learning records to determine the learner's prediction. Prints the current learning status.

Description

Learner status diagnosis device including multiple artificial neural network modules {Device for Evaluating Status of Student By Way Of Multiple Artificial Neural Networks}

The present invention relates to a learner status diagnosis device including a plurality of artificial neural network modules.

As enthusiasm for education in various languages such as Korean, English, Japanese, and Chinese heats up, parents recently began teaching children in various languages, such as Korean and English, from as early as age 4. In particular, in the case of Hangul, which is the core of early education, according to statistics from the Korea Institute for Curriculum and Evaluation, the rate of Hangul education before entering school reached 91.6%.

The existing method of teaching Korean using materials such as Korean workbooks during the care time of parents or caregivers became the mainstream, and was gradually transitioning to teaching online or offline through applications such as smartphones or tablets. . However, these Hangul education methods were either a low-cost, one-way Hangul education method based on a preset curriculum that was not related to the child's Hangul level, or an expensive, two-way Hangul education method that included guidance from a workbook teacher.

Because of this, when it was necessary to provide children's Hangul education at a low cost, even if Hangul education was conducted using an application, the difficulty level was too high or too low to arouse children's interest, and it was difficult to arouse children's interest. Alternatively, there was a problem in which caregivers had to devote a lot of teaching time to children's Korean language education. Additionally, from the perspective of parents and caregivers, there was a very difficult problem of having to choose which one was most appropriate among the plethora of options, including more than 4,000 Korean education worksheets and numerous Korean education applications available on the market.

In addition, even if an expensive two-way Hangul education method was used in which a workbook teacher visits and provides guidance, the workbook teacher usually only teaches for 15 minutes a week, and there was a problem in that parents had to take responsibility for Hangul education themselves for the rest of the time.

Therefore, the purpose of the present invention is to enable the learner's current learning status to be predicted based on the learner's learning record.

However, the problem to be solved by the present invention is not limited to the above-mentioned, and includes purposes that are not mentioned but can be clearly understood by those skilled in the art from the description below. can do.

Hereinafter, specific means for achieving the purpose of the present invention will be described.

A learner status diagnosis device including a plurality of artificial neural network modules according to an embodiment of the present invention, the plurality of artificial neural network modules include a transformer-based artificial neural network module, a front end of the transformer-based artificial neural network module, and It includes a plurality of LSTM (Long Short-Term Memory) modules installed at the rear end, and the plurality of artificial neural network modules receive learning records from the learner client, process the received learning records, and determine the learner's expected current learning state. outputs.

As described above, according to the present invention, the following effects are achieved.

According to one embodiment of the present invention, the expected current learning state can be derived based on the learner's learning record, that is, the problems the learner has solved so far.

In addition, through the learner's expected current learning state, a learning session with a level of difficulty that allows the learner to learn most interestingly and efficiently (i.e., a learning session at which the learner is expected to achieve the learning standard score) can be selected within the current curriculum. By making recommendations, it has the effect of providing customized learning sessions to learners.

However, the effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below. You will be able to.

The following drawings attached to this specification illustrate preferred embodiments of the present invention and serve to further understand the technical idea of the present invention along with the detailed description of the invention, so the present invention is limited only to the matters described in such drawings. It should not be interpreted as such.
Figure 1 is a schematic diagram showing an artificial intelligence-based learning session recommendation system according to an embodiment of the present invention.
Figure 2 is a diagram explaining the structure of a learner's learning record according to an embodiment of the present invention.
Figure 3 is a diagram explaining a method for generating a problem identifier according to another embodiment of the present invention.
Figure 4 is a diagram illustrating a process in which an artificial intelligence-based learner status diagnosis device outputs the learner's expected current learning status according to an embodiment of the present invention.
Figures 5 and 6 are diagrams illustrating the schematic structure and operating principle of the artificial neural network module of the artificial intelligence-based learner status diagnosis device according to an embodiment of the present invention.
Figure 7 is a diagram illustrating the schematic structure and operating principle of the artificial neural network module of the artificial intelligence-based reference score setting device according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating the schematic structure and operating principle of the artificial intelligence-based learning session recommendation device 3.
Figure 9 is a flowchart of a method for recommending learning sessions within a curriculum according to an embodiment of the present invention.

Hereinafter, with reference to the attached drawings, an embodiment by which a person skilled in the art can easily carry out the present invention will be described in detail. However, the present invention may be implemented in many different forms and is not limited to the examples described herein. Additionally, in explaining in detail the operating principle of a preferred embodiment of the present invention, if a detailed description of a related known function or configuration is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

In addition, the same reference numerals are used for parts that perform similar functions and actions throughout the drawings. Throughout the specification, when a specific part is said to be connected to or above/below another part, this means not only directly connected or above/below, but also indirectly through other components in between. This also includes cases where they are connected or above/below.

Furthermore, when it is said that a part "includes" a certain element throughout the specification, this does not mean excluding other elements, but may further include other elements, unless specifically stated to the contrary.

Terms such as “first” and “second” may be used to describe various components, but these components should not be limited by these terms. That is, terms are used only for the purpose of distinguishing one component from another. For example, a first component may be named a second component, and similarly, the second component may also be named a first component without departing from the scope of the present invention. The term “and/or” includes any of a plurality of related stated items or a combination of a plurality of related stated items.

Additionally, unless specifically stated otherwise, the expression “singular” is used herein to include one or more entities. Finally, as used herein, “or” means “or” non-exclusively, unless specifically stated otherwise.

In addition, the terms used in the present invention are general terms that are currently widely used as much as possible while considering the functions in the present invention, but this may vary depending on the intention or precedent of a person working in the art, the emergence of new technology, etc. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the relevant invention. Therefore, the terms used in the present invention should be defined based on the meaning of the term and the overall content of the present invention, rather than simply the name of the term.

Overview of artificial intelligence-based learning session recommendation system

Figure 1 is a schematic diagram showing an artificial intelligence-based learning session recommendation system according to an embodiment of the present invention. Referring to FIG. 1, the artificial intelligence-based learning session recommendation system 100 according to an embodiment of the present invention includes an artificial intelligence-based learner status diagnosis device (1) and an artificial intelligence-based reference score setting device (2). , and an artificial intelligence-based learning session recommendation device (3).

First, the artificial intelligence-based learner status diagnosis device 1 may receive the learning record 5 of the learner (eg, an infant) from the application module 4a of the learner client 4. At this time, the learner's learning record (5) contains problem solving information (5a) containing solutions to problems that the learner has input to date through the learning application module (4a), and which learning sessions the problems solved by the learner are. It may include learning session combination information (5b) indicating whether it is a combination. According to one embodiment of the present invention, one curriculum includes a plurality of learning sessions, and each learning session includes a plurality of problems, so the learner's problem solving information 5a also includes problems belonging to the plurality of learning sessions. An explanation can be included. Accordingly, the learning sessions of problems solved by the learner can be identified through the learning session combination information 5b.

The artificial intelligence-based learner status diagnosis device 1 can output the learner's expected current learning status based on the learning record 5 received from the application module 4a of the learner client 4. Here, the learner's expected current learning state may include, for example, whether or not correct answers are given to all questions within a plurality of learning sessions in the learner's current curriculum and an expected score for each individual learning session. The output of the learner's expected current learning state can be performed based on artificial intelligence through an artificial neural network module, which will be described in detail later.

The artificial intelligence-based standard score setting device (2) receives the learner's (e.g., infant) learning record (5) from the application module (4a) of the learner client (4) and induces the learner's efficient learning based on this. You can output the learning standard score. For example, the learning standard score is the score that can cause the learner to be most interested in learning, and may be set as the score that is expected to result in the longest learning duration. Meanwhile, in the embodiment of the present invention, it was assumed that the score expected to have the longest learning duration was the score that would cause the learner to be most interested in learning, but other factors other than the learning duration, such as the number of learning times and the number of logins, The learning standard score can also be output considering such factors. The output of these learning standard scores can be performed based on artificial intelligence through an artificial neural network module, which will be described in detail later.

The artificial intelligence-based learning session recommendation device (3) receives the learner's expected current learning state from the artificial intelligence-based learner status diagnosis device (1) and receives the learning standard score from the artificial intelligence-based standard score setting device (2). You can receive it. In addition, the artificial intelligence-based learning content recommendation device 3 can output a recommended learning session 6 based on the learner's expected current learning status and learning standard score and provide it to the learner. The recommended learning session 6 may be a learning session customized to the learner in which the learner is expected to receive a score closest to the learning standard score among a plurality of learning sessions in the current curriculum.

The artificial intelligence-based learning session recommendation system 100 according to an embodiment of the present invention is processed by a processing module of a computing device such as a specific web server, a virtual server such as a cloud server, a smartphone, a tablet PC, or a desktop PC. , may be configured to be stored in the memory module of each computing device. However, this system 100 may be distributedly processed by processing modules of two or more computing devices and may be distributed and stored in memory modules of two or more computing devices, and this is the case in the artificial intelligence-based learning session recommendation system 100. It can be designed in various ways depending on the specific implementation method.

According to one embodiment of the present invention, the expected current learning state is derived based on the learner's learning record, that is, the problems the learner has solved so far, and the difficulty level at which the learner can proceed with learning most interestingly and efficiently (i.e. By recommending a learning session within the current curriculum (a learning session in which the learner is expected to obtain a learning standard score), it has the effect of providing a customized learning session to the learner.

Artificial intelligence-based learner status diagnosis device

Figure 2 is a diagram explaining the structure of a learner's learning record according to an embodiment of the present invention.

Referring to FIG. 2, the learner's current learning record 5 includes a set of problem identifiers (e ₁₁ , e ₂₂ ,... , e _mn ), the learner's problem solving information for each problem (r ₁₁ , r ₂₂ ,..., r _mn ), and learning session combination information (5b) indicating the combination of learning sessions to which the problems solved by the learner belong. ) may include. For example, if the problem identifier e ₁₁ belongs to learning session 1, the problem identifier e ₂₂ belongs to learning session 2, and the problem identifier e _mn belongs to learning session m, the learning session combination information 5b is the learner's learning information. Record (5) can indicate that learning sessions 1, 2, and m are combined.

Meanwhile, the learner's learning record 5 may include not only the current learning record for the current curriculum but also learning records for curricula previously solved by the learner. In this case, learning records 5 may be sorted by curriculum (eg, by curriculum ID) and/or by date. In addition, the database stored on the server of the learning application platform may store multiple learners' learning records (5) by learner ID (e.g., profile ID), and in addition, the learning records (5) along with the mode Information (indicating whether the problem is a learning-type problem or a query-type problem), etc. may be added and stored, but this is not limited.

According to one embodiment of the present invention, each problem identifier (e) may be composed of a combination of a game type (g, game type) and a learning target (t, target). That is, the problem identifiers e ₁ =g ₁ t ₁ , e ₂ =g ₂ t ₂ , […] ], e _k-1 =g _k-1 t _k-1 . Here, the game type (g) indicates what type of game the problem is being presented as. For example, the game type (g) is a "sound learning" game that learns the shapes and sounds of letters, and compares and discriminates with other letters. It can include a "discrimination" game that involves adding consonants, vowels, and consonants, a "combining" game that involves adding consonants, vowels, and final consonants, a "tracing" game that teaches you how to write letters correctly, and a "speaking" game that trains you to listen and speak. Each game can include Type (g) can be indicated by an identifier or label. Additionally, the learning object (t) may be a letter, word, consonant, vowel, or phoneme to be learned, and each learning object (t) may be displayed as an identifier or label.

Figure 3 is a diagram explaining a method for generating a problem identifier according to another embodiment of the present invention. Referring to Figure 3, according to another embodiment of the present invention, each problem identifier (e) may be composed of a combination of a game type (g, game type) and a learning target (t, target), where the game type ( g) inputs the game screen into a CNN (Convolution Neural Network) encoder and determines the game vector as the output, and the learning target (t) inputs the learning target text into the CNN encoder and determines it as the target vector as the output. It may be determined, and each problem identifier (e) may be composed of a combination of these game vectors and target vectors.

Figure 4 is a diagram illustrating a process in which an artificial intelligence-based learner status diagnosis device outputs the learner's expected current learning status according to an embodiment of the present invention.

Referring to FIG. 4, the learner of the learner client 4 can input solutions to problems of learning sessions in the current curriculum through the learner client 4, and this can become the learner's learning record 5. . According to the example shown in Figure 4, the learning record 5 is the learner's problem solving for the problems solved so far (i.e., each problem identifier e ₁₁ , e ₂₂ , [...], e _mn ) Information (5a) (r ₁ , r ₂ , [...], r _t ) and which learning session in the category each problem (e ₁₁ , e ₂₂ , [...], e _mn ) belongs to. It may include learning session combination information (5b). These learning records (5) can be transmitted from the learner client (4) to the artificial intelligence-based learner status diagnosis device (1).

The artificial intelligence-based learner status diagnosis device 1 can output the learner's expected current learning status for all problems in multiple learning sessions within a category based on the received learning record 5. Here, the learner's expected current learning state (a ₁₁ , a ₁₂ ... a ₂₁ , a ₂₂ , a _m1 , a _m2 , [...] a _mn ) is given a value of "0" for the correct answer, A value of "1" may be given for an incorrect answer, and a value of "2" indicating completion of learning may be given. Alternatively, for learning-complete problems, when learning is completed, a value of "0" may be assigned as in the correct answer.

In addition, the artificial intelligence-based learner status diagnosis device 1 may output the learner's expected current learning status (i.e., whether the questions are answered incorrectly) in the form of a probability value. For example, if the learner's expected current learning state is (a ₁₁ = 0.34, a ₁₂ = 0.78 [...], a ₂₁ = 0.88, a ₂₂ = 0.73, a _m1 = 0.55, a _m2 = 0.21 ... a It can be output as a probability value, such as _mn = 0.98). In this case, based on a predetermined standard threshold, for example, 0.607, if the probability value is 0.607 or more, the question is considered to be answered correctly, and if the probability value is less than 0.607, the question is considered to be an incorrect answer. can be set. These reference thresholds are the probability of predicting an actual correct answer as a correct answer (TPR, true positive rate), the probability of predicting an actual incorrect answer as a correct answer (TNR, true negative rate), the probability of predicting an actual incorrect answer as a correct answer (FPR, false positive rate), It may be determined by considering the probability of predicting the actual correct answer as an incorrect answer (false negative rate (FNR)), but is not limited thereto.

According to one embodiment, as shown in Figure 4, the learner's expected current learning state only includes whether correct answers (and learning completion) to all problems within a plurality of learning sessions in the curriculum as described above Alternatively, it may further include expected scores for individual study sessions. For example, the expected score for the learning session may be calculated based on whether or not incorrect answers are output by the artificial intelligence-based learner status diagnosis device 1 for problems within the learning session.

Figures 5 and 6 are diagrams illustrating the schematic structure and operating principle of the artificial neural network module of the artificial intelligence-based learner status diagnosis device according to an embodiment of the present invention.

Referring to FIGS. 5 and 6, according to an embodiment of the present invention, the artificial neural network module 10 of the artificial intelligence-based learner status diagnosis device 1 includes the first block 10a of the transformer-based artificial neural network module. It may include a structure in which LSTM (Long Short-Term Memory) modules 41, 42, and 43 are added to the front and rear ends of (e.g., encoder) and the second block 10b (e.g., decoder). there is. This structure has the effect of improving the output accuracy of the learner's expected current learning state through the artificial intelligence-based learner state diagnosis device (1).

First, the operating principle of the pre-learning session of the artificial neural network module 10 of the artificial intelligence-based learner status diagnosis device 1 of FIG. 5 will be briefly described.

In the pre-learning session, pre-learning learning records including learning records previously performed by a plurality of learners are input to the artificial neural network module 10, and the input learning records for pre-learning are processed to produce the artificial neural network module 10. The expected learning state of each of the plurality of learners is output, and the expected current learning state of each of the plurality of learners is compared with the actual learning result of each of the plurality of learners, so that the error between the expected current learning state and the actual learning result is calculated. The artificial neural network module 10 can be pre-trained by constructing a loss function in a direction that is minimized.

More specifically, in the pre-training session, the problem identifier (e ₁ , e ₂ , [...], e _{k-1) of the learning record 5 for pre-training is stored in the first block 10a of the artificial neural network module 10} . can be input in time series. The learning record 5 for prior learning may include learning records previously performed by a plurality of learners. At this time, as described above, each problem identifier (e ₁ , e ₂ , [...], e _k-1 ) may be composed of a combination of a game type (g, game type) and a learning target (t, target). That is, the problem identifiers e ₁ =g ₁ t ₁ , e ₂ =g ₂ t ₂ , […] ], e _k-1 =g _k-1 t _k-1 . Meanwhile, according to the example of Figure 3, each problem identifier (e) input in time series is composed of a combination of a game type (g, game type) and a learning target (t, target), where the game type (g) is a game The screen is input to a CNN (Convolution Neural Network) encoder and determined as the game vector as the output, and the learning target (t) can be input to the CNN encoder and determined as the target vector as the output. Each problem identifier (e) may be composed of a combination of these game vectors and target vectors.

In addition, the problem identifier input in time series in this way is input to the embedding module of the first block 10a and vectorized, so that embedding information can be output and input to the LSTM module 41. The LSTM module 41 calculates how much past content will be forgotten or remembered based on the information at the current time, adds current information to the result, and transmits the information to the next time. The LSTM module 41 ) based on the embedding information, linearly embeds a bundle of time series encoding information through different linear layers to form Q (Query feature), K (Key feature), and V (Value feature) including feature dimensions, and then Q ( Query feature), K (Key feature), and V (Value feature) can be used as input data for the Multi-head Attention Layer (i.e., Multi-head Attention with upper triangular mask). Q, K, and V can be generated from embedding information through different linear embeddings. Q and K are input as input data of the 1st MatMul operation, calculate the similarity for all K with respect to Q through scale and softmax operation, and output a similarity vector between a series of embedding information, and this similarity vector and V are the 2nd MatMul It can be entered as input data for the operation. Additionally, the Multi-head Attention Layer can output attention information based on the input Q, K, and V values and input it to the Feed Forward Layer. These Multi-head Attention Layers and Feed Forward Layers can be repeated, for example, six times, and the repetition can be done in such a way that the encoded information output from the previous Feed Forward Layer becomes the input to the next Multi-head Attention Layer. And the last Feed Forward Layer can output the final encoding information, and this final encoding information is transmitted to the second multi-head attention layer of the second block 10b of the artificial neural network module 10. ) can be entered as a value.

In addition, in the pre-learning session, the second block 10b of the artificial neural network module 10 includes learning session combination information 5b and problem solving information 5a (s, r ₁ , r ₂ , [...], r _{k -1} ) can be entered. Here, s in the problem solving information 5a may be a start signal, and the learner's learning duration can be determined by comparing this start signal with the learning time of the last problem in the learning record. In this way, the learning session combination information (5b) and problem solving information (5a) (s, r ₁ , r ₂ , [...], r _k-1 ) input to the second block (10b) are the same as those of the second block (10b). By being input to the embedding module and vectorized, the embedding information can be output and input to the LSTM module 42. The LSTM module 42 linearly embeds a bundle of time series encoding information through different linear layers based on the embedding information, and configures it into Q (Query feature), K (Key feature), and V (Value feature) including feature dimensions. Afterwards, Q (Query feature), K (Key feature), and V (Value feature) can be used as input data for the first Multi-head Attention Layer (i.e., Multi-head Attention with upper triangular mask). And the first Multi-head Attention Layer can output a Q (Query) value, and this Q value is input to the second Multi-head Attention Layer. The second multi-head attention layer can receive the final encoding information from the first block 10a as K and V values. And the second multi-head attention layer can output attention information based on the input Q, K, and V values and input it to the feed forward layer. These first Multi-head Attention Layer, second Multi-head Attention Layer, and Feed Forward Layer may be repeated, for example, six times, and the decoding information output from the previous Feed Forward Layer is transmitted to the next first Multi-head Attention Layer. Repetition can be done in a way that inputs . And the last Feed Forward Layer can output the final decoding information. This final decoding information is input to the LSTM module 43, and the time series decoding information output by the LSTM module 43 passes through the Linear Layer to the learner's expected current learning state. It can be output as (a ₁ , a ₂ , […], a _k-1 ).

And, the learner's expected current learning state (a ₁ , a ₂ , [...], a _k-1 ) output from the artificial neural network module 10 is the learner's actual learning results (a' ₁ , a' ₂ , [...] ], a' _k-1 ), and as a result of the comparison, the loss function is configured in a direction that minimizes the error or loss between the two values (for example, Binary Cross Entropy method), so that the artificial neural network module 10 Can be pre-trained. In other words, the artificial neural network module 10 predicts the current learning state (a ₁ , a ₂ , […], a _k-1 ) and the learner’s actual learning results (a’ ₁ , a’ ₂ , […], a’ A method of updating the weight of each layer of the artificial neural network 10 based on the loss function between _(k-1 ) can be adopted.

However, the specific implementation method of the transformer-LSTM based artificial neural network module as described above may vary in various ways and is not limited thereto.

Referring to FIG. 6, the operating principle of the inference session of the artificial neural network module 10 of the artificial intelligence-based learner status diagnosis device 1 will be described. Referring to FIG. 6, problem identifiers (e ₁₁ , By inputting e ₂₁ , […], e _mn ) and inputting the learning session combination information 5b and the start signal s into the second block 10b, the artificial neural network module 10 determines the learner's expected current learning. Status (a ₁₁ , a ₂₁ , […], a _mn ) can be output. The learner's expected current learning status may include, for example, correct answers to all questions within a plurality of learning sessions in the current curriculum and an expected score for each individual learning session.

Artificial intelligence-based standard score setting device

Figure 7 is a diagram illustrating the schematic structure and operating principle of the artificial neural network module of the artificial intelligence-based reference score setting device according to an embodiment of the present invention.

Referring to FIG. 7, in the pre-learning session of the artificial neural network module of the artificial intelligence-based standard score setting device (2), learning session combination information (5b) and problem solving of the learning records (5) of multiple learners for pre-learning Information 5a (s, r ₁ , r ₂ , […], r _k-1 ) may be input. Here, the problem solving information 5a includes a start signal s, and the learner's learning duration can be known through this start signal s and the solution information r _k-1 of the last problem in the learning record.

Afterwards, the learning session combination information 5b and the problem solving information 5a may be output as first embedding information and second embedding information, respectively, through the embedding module. And, the first embedding information and the second embedding information may be input to the LSTM module. However, according to another example, other artificial neural network modules such as transformer module, CNN module, RNN module, multi-perceptron module, GAN module, etc. may be used instead of the LSTM module, but the present invention is not limited thereto.

The LSTM module can output vector information based on the first embedding information and the second embedding information, and this vector information can be output as an expected learning standard score through the linear module. Here, the learning standard score is the score that makes the learner most interested in learning, and can be set as the score that is expected to result in the longest learning duration. In addition, the expected learning standard score in the pre-learning session can be compared with the actual maximum learning time score, which was the longest learning duration for each of the multiple learners, and as a result of the comparison, there is a difference between the expected learning standard score and the actual maximum learning time score. The artificial neural network module can be pre-trained by constructing a loss function in a direction that minimizes errors or losses (for example, L1 or L2 norm method). In other words, the artificial neural network module of the artificial intelligence-based standard score setting device (2) adopts a method of updating the weight of each layer of the artificial neural network based on the error or loss function between the expected learning standard score and the actual maximum learning time score. can do.

In addition, the artificial neural network module for which the pre-learning session of the artificial intelligence-based standard score setting device (2) has been completed provides the learner's learning session combination information (5b) and problem solving information (5a) (s, r ₁ , r) in the inference session. ₂ , […], r _k-1 ) can be input to output the learning standard score that can generate the greatest interest of a specific learner (i.e., the learner's learning time is expected to be the longest).

Artificial intelligence-based learning session recommendation device

FIG. 8 is a diagram illustrating the schematic structure and operating principle of the artificial intelligence-based learning session recommendation device 3.

Referring to Figure 8, the artificial intelligence-based learning session recommendation device (3) receives the learner's expected current learning state from the artificial intelligence-based learner status diagnosis device (1), and the artificial intelligence-based reference score setting device (2) ) You can receive learning standard scores from. In addition, the artificial intelligence-based learning content recommendation device (3) outputs a recommended learning session (6) based on the learner's expected current learning status and learning standard score and provides it to the learner's client (4). .

According to a non-limiting example, the artificial intelligence-based learning content recommendation device 3 includes the learning standard score received from the standard score setting device 2 and the expected current learning state of the learner for each learning session in the curriculum. The scores can be compared, and the learning session with the expected score closest to the learning standard score among the learning sessions in the curriculum can be determined as the recommended learning session (6). According to the example in Figure 8, the learner's expected score for learning session 1 is 63 points, the learner's expected score for learning session 2 is 41 points, and the learner's expected score for learning session m is 84 points. If the learning standard score received from the device 2 is 60, learning session 1 closest to the learning standard score of 60 may be determined as the recommended learning session.

However, the example in FIG. 8 assumes that, in addition to learning sessions 2 and m, there are no learning sessions whose expected scores are closer to 60 than learning session 1, and according to another example, there are more than two learning sessions whose expected scores are closest to the standard score. This can be. In this case, a specific learning session may be determined as a recommended learning session according to a predetermined rule. For example, among learning sessions whose expected score is closest to the standard score, a learning session adjacent to a learning session recently studied by the learner may be determined as a recommended learning session.

The recommended learning session (6) determined in this way is a customized learning session that is expected to result in the learner receiving the score closest to the learning standard score among the plurality of learning sessions in the current curriculum, that is, that is most likely to arouse the learner's interest. It could be a learning session.

Recommended methods for learning sessions within the curriculum

Figure 9 is a flowchart of a method for recommending learning sessions within a curriculum according to an embodiment of the present invention.

Referring to FIG. 9, first, in step S910, the learner's expected current learning status for problems of a plurality of learning sessions in the current curriculum may be output based on the learning record received from the learner client. At this time, the learner's expected current learning state may include whether or not incorrect answers are given to questions of a plurality of learning sessions in the current curriculum and an expected score for each individual learning session.

Here, the learning record may include the learner's learning record for problems within the current curriculum. In addition, the learning record includes problem solving information that includes solutions to problems that the learner has entered through the learning application module to date, and learning session combination information that indicates what kind of combination of learning sessions the problems solved by the learner consisted of. Wow, it can contain a set of problem identifiers for the problems solved by the learner. Additionally, each problem identifier in the problem identifier set may be composed of a combination of the game type and learning target of the problem corresponding to the problem identifier.

In step S920, a learning standard score expected to result in the longest learning duration of the learner may be output based on the learner's learning record. In the embodiment of the present invention, it was assumed that the score expected to have the longest learning duration was the score that would cause the learner to be most interested in learning, but other factors other than the learning duration, such as the number of learning times, the number of logins, etc. Taking this into account, the learning standard score can also be output.

Next, in step S930, based on the expected score of the learner's expected current learning state and the learning standard setting score, a learning session among the plurality of learning sessions in the current curriculum may be determined as a recommended learning session.

computer-readable recording medium

It is obvious that each step or operation according to the embodiments of the present invention can be performed by a computer including one or more processors by executing a computer program stored in a computer-readable recording medium.

Each instruction stored in the above-described recording medium can be implemented through a computer program programmed to perform each corresponding step, and such computer program may be stored in a computer-readable recording medium and can be executed by a processor. do. A computer-readable recording medium may be a non-transitory readable medium. At this time, a non-transitory readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as registers, caches, and memories. Specifically, programs for performing the various methods described above include semiconductor memory devices such as erasable programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), and flash memory devices, internal hard disks, and removable memory devices. It may be provided and stored in non-transitory readable media such as magnetic disks such as disks, optical-magnetic disks, and non-volatile memory including CD-ROM and DVD-ROM disks.

Methods according to various examples disclosed in this document may be provided and included in a computer program product. The computer program product may be distributed in the form of a machine-readable storage medium (e.g. compact disc read only memory (CD-ROM)) or online through an application store (e.g. Play Store ^™ ). In the case of online distribution, at least a portion of the computer program product may be at least temporarily stored or created temporarily in a storage medium such as the memory of a manufacturer's server, an application store's server, or a relay server.

As described above, a person skilled in the art to which the present invention pertains will understand that the present invention can be implemented in other specific forms without changing its technical idea or essential features. Therefore, the above-described embodiments should be understood in all respects as illustrative and not restrictive. The scope of the present invention is indicated by the claims described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and the equivalent concept should be construed as being included in the scope of the present invention.

The features and advantages described herein are not all-inclusive, and many additional features and advantages will become apparent to those skilled in the art upon consideration of the drawings, specification, and claims. Moreover, it should be noted that the language used herein has been selected primarily for readability and teaching purposes and may not be selected to describe or limit the subject matter of the invention.

The foregoing description of embodiments of the invention has been presented for purposes of illustration. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Those skilled in the art will appreciate that many modifications and variations are possible in light of the above disclosure.

Therefore, the scope of the present invention is not limited by the detailed description, but by any claims of the application based thereon. Accordingly, the disclosure of embodiments of the invention is illustrative and does not limit the scope of the invention as set forth in the claims below.

100: Artificial intelligence-based learning session recommendation system
1: Artificial intelligence-based learner status diagnosis device
2: Artificial intelligence-based standard score setting device
3: Artificial intelligence-based learning session recommendation device
4: Learner Client
4a: Application module
5: Learning record
5a: Problem solving information
5b: Learning Session Combination Information
6: Recommended Learning Sessions

Claims (1)

A learner status diagnosis device including a plurality of artificial neural network modules,
The plurality of artificial neural network modules include a transformer-based artificial neural network module and a plurality of Long Short-Term Memory (LSTM) modules installed at the front and rear ends of the transformer-based artificial neural network module,
The plurality of artificial neural network modules receive learning records from the learner client, process the received learning records, and output the learner's expected current learning state.
A learner status diagnosis device including a plurality of artificial neural network modules.