Method for predicting response accuracy, question recommendation method and device

Abstract

The present disclosure provides a prediction method for answer accuracy, a question recommendation method and device, the prediction method for answer accuracy comprising: obtaining first question information and first time information corresponding to answered questions of a target user, and second question information and second time information corresponding to to-be-answered questions; determining forgetting information between the answered questions and the to-be-answered questions according to the first question information, the first time information, the second question information and the second time information; determining similarity information between the answered questions and the to-be-answered questions according to the first question information and the second question information; and predicting answer accuracy of the target user for the to-be-answered questions based on the forgetting information and the similarity information. The prediction result obtained by the present disclosure is more in line with the actual situation, and can effectively improve the accuracy of the prediction result. On this basis, the rationality of the title of the question recommended for the user can be effectively guaranteed.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to methods and apparatus for predicting answer accuracy, recommending questions, and related devices. Background Technology

[0002] Knowledge tracing technology can predict a student's future performance based on their past performance, such as the probability of accurately answering the next question. This can provide personalized tutoring or question recommendations to improve students' learning efficiency. However, the inventors discovered that the predictions obtained using existing knowledge tracing technology are inaccurate, leading to unreasonable question recommendations based on inaccurate predictions. Summary of the Invention

[0003] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a method for predicting answer accuracy, a method for recommending questions, and an apparatus.

[0004] According to one aspect of this disclosure, a method for predicting answer accuracy is provided, comprising: acquiring first question information and first time information corresponding to questions already answered by a target user, and second question information and second time information corresponding to questions to be answered; determining forgetting information between the already answered questions and the questions to be answered based on the first question information, the first time information, the second question information, and the second time information; determining similarity information between the already answered questions and the questions to be answered based on the first question information and the second question information; and predicting the target user's answer accuracy for the questions to be answered based on the forgetting information and the similarity information.

[0005] According to another aspect of this disclosure, a question recommendation method is provided, comprising: acquiring a plurality of questions to be predicted; acquiring the answer accuracy rate of a target user for each of the questions to be predicted; wherein the answer accuracy rate is determined based on the aforementioned prediction method for answer accuracy rate; and recommending target questions to the target user based on the answer accuracy rate of each of the questions to be predicted.

[0006] According to another aspect of this disclosure, an apparatus for predicting answer accuracy is provided, comprising: an information acquisition module, configured to acquire first question information and first time information corresponding to a question answered by a target user, and second question information and second time information corresponding to a question to be answered; a forgotten information determination module, configured to determine forgotten information between the answered question and the question to be answered based on the first question information, the first time information, the second question information, and the second time information; a similarity information determination module, configured to determine similarity information between the answered question and the question to be answered based on the first question information and the second question information; and an accuracy prediction module, configured to predict the answer accuracy of the target user for the question to be answered based on the forgotten information and the similarity information.

[0007] According to another aspect of this disclosure, a question recommendation device is provided, comprising: a question acquisition module for acquiring a plurality of questions to be predicted; an accuracy acquisition module for acquiring the answer accuracy rate of a target user for each of the questions to be predicted; wherein the answer accuracy rate is determined based on the answer accuracy prediction method described in any one of the claims; and a question recommendation module for recommending target questions to the target user based on the answer accuracy rate of each of the questions to be predicted.

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: a processor; and a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the prediction method for answer accuracy, or to perform the question recommendation method.

[0009] According to another aspect of this disclosure, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the aforementioned method for predicting answer accuracy, or for performing the aforementioned method for recommending questions.

[0010] The method and apparatus for predicting the accuracy of answers provided in this embodiment of the present disclosure fully consider the phenomenon of user forgetting and the impact of question similarity on the accuracy of answering. Therefore, it can obtain the forgetting information and similarity information between the already answered questions and the questions to be answered, and combine the forgetting information and similarity information to predict the user's accuracy of answering, so that the prediction results are more in line with the actual situation and help to further improve the accuracy of the prediction results.

[0011] The question recommendation method and apparatus provided in this embodiment can use the aforementioned method for predicting the accuracy of a user's answers to multiple questions to make question recommendations, thereby effectively ensuring the rationality of the target questions recommended to the user.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0014] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating a method for predicting the accuracy of responses provided in an embodiment of this disclosure;

[0016] Figure 2 A flowchart illustrating a method for predicting the accuracy of responses provided in an embodiment of this disclosure;

[0017] Figure 3 A flowchart illustrating a topic recommendation method provided in an embodiment of this disclosure;

[0018] Figure 4 A schematic diagram of the structure of a device for predicting the accuracy of responses provided in an embodiment of this disclosure;

[0019] Figure 5 This is a schematic diagram of the structure of a topic recommendation device provided in an embodiment of the present disclosure;

[0020] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0021] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0022] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0023] The term "comprising" and its variations as used in this disclosure are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc., mentioned in this disclosure are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0024] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0025] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0026] Figure 1 This is a flowchart illustrating a method for predicting answer accuracy according to an embodiment of this disclosure. The method can be executed by an answer accuracy prediction device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. Figure 1 As shown, the method mainly includes the following steps S102 to S108:

[0027] Step S102: Obtain the first question information and first time information corresponding to the questions already answered by the target user, as well as the second question information and second time information corresponding to the questions to be answered.

[0028] In practical applications, assuming the question to be answered is question t+1, the questions already answered can be question t, questions 1 through t, or N individual questions preceding question t+1. The specific settings can be flexibly configured according to the actual situation. Furthermore, this embodiment does not limit the first question information. For example, the first question information can include the text information of the already answered questions, or it can include the knowledge point information of the already answered questions. The knowledge point information can further include main knowledge point information and secondary knowledge point information, etc. The second question information is similar and will not be elaborated here. The first time information can be the answering time of the already answered questions. Specifically, it can include the start time of answering the already answered questions, or it can include the end time of answering or the answering duration, etc., without limitation. The time information of the question to be answered can be the start time of the question to be answered, but since the question to be answered has not yet been completed, it is necessary to predict whether the answer to the question to be answered is accurate.

[0029] Step S104: Based on the first question information, the first time information, the second question information, and the second time information, determine the forgotten information between the answered questions and the unanswered questions.

[0030] In this embodiment, the possibility of user forgetting is fully considered, and forgetting can affect prediction results. Therefore, this embodiment obtains forgetting information based on the question information and time information of both answered and unanswered questions. This forgetting information can be used to describe the target user's forgetting of knowledge points relative to answered questions when faced with unanswered questions, thereby measuring the user's degree of forgetting. In practical applications, the first question information includes first knowledge point information, which can further indicate the main and secondary knowledge points contained in the answered questions. The second question information includes second knowledge point information, which can further indicate the main and secondary knowledge points contained in the unanswered questions. The type and number of knowledge points hits in the unanswered questions relative to the answered questions are further obtained. Combined with a preset forgetting weight, the user's forgetting of the hit knowledge points is comprehensively measured.

[0031] Step S106: Based on the information of the first question and the information of the second question, determine the similarity information between the questions that have been answered and the questions that need to be answered.

[0032] In this embodiment, the impact of question similarity on the prediction of answer accuracy is fully considered. Therefore, the similarity between questions is obtained based on question information. This embodiment does not limit the method of obtaining question similarity. For example, the similarity information may include question text similarity and / or knowledge point similarity, and may also include answer result similarity, etc. Specifically, a large number of historical user answers to questions that have been answered and questions that are yet to be answered can be obtained to evaluate the similarity of the answer results between the answered and unanswered questions. A comprehensive evaluation is then performed based on the similarity of the answered and unanswered questions in multiple dimensions such as text, knowledge points, and answer results to obtain the comprehensive similarity between the answered and unanswered questions. The comprehensive similarity obtained through the above method can fully cover the similarity evaluation results of each dimension and match the answered and unanswered questions in multiple dimensions, effectively ensuring the accuracy and reliability of the similarity evaluation between the answered and unanswered questions. The above is only an illustrative example. In practical applications, the similarity information may only include similarity in one dimension. The specific details can be flexibly determined according to the actual situation and are not limited here.

[0033] Step S108: Based on forgotten information and similar information, predict the accuracy rate of the target user's answers to the questions to be answered.

[0034] In some specific implementation examples, forgotten information and similar information can be fused to obtain fused information. Based on the fused information, the accuracy rate of the target user's answers to the questions to be answered can be predicted. That is, when making accuracy prediction, both forgotten factors and question similarity factors are taken into account, so that the prediction results are more accurate.

[0035] In summary, the above-mentioned method for predicting the accuracy of answers provided in this embodiment of the present disclosure fully considers the phenomenon of user forgetting and the impact of question similarity on the accuracy of answers. Therefore, it can obtain forgetting information and similarity information between the already answered questions and the questions to be answered, and combine the forgetting information and similarity information to predict the user's accuracy of answers, so that the prediction results are more in line with the actual situation and help to further improve the accuracy of the prediction results.

[0036] Considering that some questions may contain multiple knowledge points, including both primary and secondary knowledge points, such as "In the linear function y = kx + b, y decreases as x increases, and b < 0, then the graph of this function does not pass through which quadrant?", the primary knowledge point is the monotonicity of linear functions, and the secondary knowledge point is the graph of the function. Therefore, this embodiment further subdivides the knowledge points. The first question information includes the first primary knowledge point information and the first secondary knowledge point information, and the second question information includes the second primary knowledge point information and the second secondary knowledge point information. The above knowledge point information can be obtained directly from the knowledge point annotation information of the question in the question bank, or it can be obtained by knowledge point identification and analysis of the question. There is no restriction here. Based on this, the above step S104, that is, the step of determining the forgotten information between the answered questions and the unanswered questions based on the first question information, the first time information, the second question information, and the second time information, can be performed with reference to the following steps A and B:

[0037] Step A: Based on the first primary knowledge point information, the first secondary knowledge point information, the second primary knowledge point information, and the second secondary knowledge point information, determine the knowledge point hit result of the question to be answered relative to the already answered questions.

[0038] In some implementation examples, the knowledge point matching results include: the number of times the second primary knowledge point matches the first primary knowledge point, the number of times the second primary knowledge point matches the first secondary knowledge point, the number of times the second secondary knowledge point matches the first primary knowledge point, and the number of times the second secondary knowledge point matches the first secondary knowledge point. For example, suppose the answered questions contain 2 primary knowledge points (knowledge points A and B) and 4 secondary knowledge points (knowledge points C, D, E, and F), and the question to be answered contains 1 primary knowledge point (knowledge point D) and 3 secondary knowledge points (knowledge points A, E, and M). Then, the number of times the second primary knowledge point matches the first primary knowledge point is 0, the number of times the second primary knowledge point matches the first secondary knowledge point is 1 (i.e., knowledge point D), the number of times the second secondary knowledge point matches the first primary knowledge point is 1 (i.e., knowledge point A), and the number of times the second secondary knowledge point matches the first secondary knowledge point is 1 (i.e., knowledge point E).

[0039] Step B involves determining the forgetting information between answered and unanswered questions based on the knowledge point hit results, first-time information, and second-time information. It's understandable that the longer the time interval, the greater the degree of forgetting. The knowledge point hit results can effectively measure the types of knowledge point hits (such as the hit rate between primary and secondary knowledge points) for unanswered questions relative to answered questions, as well as the quantity of each type of hit. This effectively reflects the relationship between unanswered and answered questions. Combining the above information, it's possible to objectively assess the forgetting level of users when faced with unanswered questions after previously encountering answered ones.

[0040] In some implementation examples, step B can be performed as follows: steps B1 to B3:

[0041] Step B1: Based on the first time information and the second time information, determine the time difference information between the answered questions and the questions to be answered. In some implementation examples, the first time information is the start time of the answered questions, and the second time information is the start time of the questions to be answered. Specifically, it can be represented by the completion time of the previous question of the questions to be answered, or by the time when the user triggers the questions to be answered. There are no restrictions here. In this case, the time difference information includes the time difference between the start time of the questions to be answered and the start time of the answered questions. In addition, if the answered questions contain multiple knowledge points, and the multiple knowledge points are examined separately, such as a question containing multiple sub-questions, the first time information can include the start time of each knowledge point of the answered questions. In this case, the time difference information includes the time difference between the start time of each knowledge point of the questions to be answered and the start time of the answered questions. In addition, if multiple knowledge points in the answered questions are combined and tested together, the start time of answering each knowledge point is considered to be the start time of answering the answered questions. In this case, the time difference information can still only include the time difference between the start time of the question to be answered and the start time of the answered questions.

[0042] Step B2: Obtain multiple preset forgetting weights; wherein, the multiple forgetting weights include: the first forgetting weight corresponding to the second primary knowledge point hitting the first primary knowledge point, the second forgetting weight corresponding to the second primary knowledge point hitting the first secondary knowledge point, the third forgetting weight corresponding to the second secondary knowledge point hitting the first primary knowledge point, and the fourth forgetting weight corresponding to the second secondary knowledge point hitting the first secondary knowledge point.

[0043] This embodiment of the disclosure uses forgetting weights to characterize the impact of knowledge point hit results on user forgetting. Different knowledge point hit types can have different forgetting weights. For example, considering that primary knowledge points leave a deeper impression on users, users tend to concentrate on these primary knowledge points when answering questions. Therefore, the forgetting weight corresponding to a second primary knowledge point hitting a first primary knowledge point may be less than the forgetting weight corresponding to a second primary knowledge point hitting a first secondary knowledge point. The above is merely an illustrative example and should not be considered a limitation. In practical applications, the magnitude of the forgetting weights can be flexibly set.

[0044] Step B3: Based on the time difference information, multiple forgetting weights, and knowledge point hit results, determine the forgetting information between answered questions and unanswered questions. For example, the knowledge point hit results include various types of knowledge point hit counts, including: the number of times the second main knowledge point hits the first main knowledge point, the number of times the second main knowledge point hits the first secondary knowledge point, the number of times the second secondary knowledge point hits the first main knowledge point, and the number of times the second secondary knowledge point hits the first secondary knowledge point. Therefore, a weighted summation can be performed based on multiple forgetting weights and the knowledge point hit counts corresponding to each forgetting weight. Combined with the time difference information, a preset forgetting algorithm is used to comprehensively determine the forgetting information between answered questions and unanswered questions. The forgetting information can be represented in the form of forgetting values.

[0045] In some implementation examples, considering that the questions to be answered may contain knowledge points that were not covered by the already answered questions, the already answered questions have no impact on the questions to be answered. Since the user did not encounter these knowledge points when answering the already answered questions, it can be considered that the user has completely forgotten these knowledge points when facing the questions to be answered, and needs to rethink these knowledge points. In order to make the forgetting information more reliable and comprehensive, and to more objectively present the user's forgetting phenomenon, this embodiment of the disclosure also incorporates information on these knowledge points into the forgetting information, specifically referring to the following steps B3.1 to B3.3:

[0046] Step B3.1: Obtain the first penalty item and the second penalty item; wherein, the first penalty item is determined based on the number of times the first primary knowledge point is not hit by the second primary knowledge point and the second secondary knowledge point and the preset first penalty coefficient; the second penalty item is determined based on the number of times the first secondary knowledge point is not hit by the second primary knowledge point and the second secondary knowledge point and the preset second penalty coefficient.

[0047] For example, suppose the answered questions contain 2 main knowledge points (knowledge points A and B) and 4 secondary knowledge points (knowledge points C, D, E, and F), and the questions to be answered contain 1 main knowledge point (knowledge point D) and 3 secondary knowledge points (knowledge points A, E, and M). Then, the number of times the first main knowledge point is not hit by the second main knowledge point and the second secondary knowledge point is 1 (knowledge point B), and the number of times the first secondary knowledge point is not hit by the second main knowledge point and the second secondary knowledge point is 2 (knowledge points C and F). The penalty coefficient can be used to set the degree of influence of the number of knowledge points not hit on the forgotten information. The above coefficient can be flexibly set according to the actual situation, such as being set to a positive value, a negative value, or a zero value (i.e., having no effect), without any specific restrictions.

[0048] Step B3.2 involves performing a weighted average based on time difference information, multiple forgetting weights, and knowledge point hit results to determine the comprehensive forgetting item.

[0049] In some implementation examples, it is assumed that the start time of answering each knowledge point in the already answered questions is the start time of answering the already answered questions. The time difference information includes the time difference between the start time of the question to be answered and the start time of the already answered questions. The knowledge point hit results include the number of hits of various knowledge points for the question to be answered relative to the already answered questions. Then, a weighted average can be first performed based on multiple forgetting weights and the number of hits of various knowledge points for the question to be answered relative to the already answered questions. Then, the weighted average result is multiplied by the time difference to obtain the result value. The negative value of the result value is then used as the exponent of the natural constant e to obtain the comprehensive forgetting term.

[0050] Step B3.3: Based on the comprehensive forgetting item, the first penalty item, and the second penalty item, determine the forgetting information between the answered questions and the unanswered questions.

[0051] For example, the sum of the forgetting term, the first penalty term, and the second penalty term can be used as forgetting information between the answered question and the unanswered question. For instance, assuming the answered question is the i-th question, the forgetting information between the answered question and the unanswered question is:

[0052]

[0053] in, l1λ1 is the comprehensive forgetting term; l2λ2 is the first penalty term; l2λ2 is the second penalty term.

[0054] Specifically, the letters in the above formula have the following meanings:

[0055] w zz This indicates the forgetting weight of the main knowledge points in the unanswered question matching the main knowledge points in the already answered questions, which is also the aforementioned first forgetting weight; wzf This indicates the forgetting weight of the main knowledge point in the question to be answered hitting the secondary knowledge point in the already answered question, which is also the aforementioned second forgetting weight; w fz This indicates the forgetting weight of secondary knowledge points in the unanswered question that match the main knowledge points in the already answered question; this is also known as the aforementioned third forgetting weight. ff This represents the forgetting weight of secondary knowledge points in the question to be answered and secondary knowledge points in the already answered question, which is also the aforementioned fourth forgetting weight; △t represents the time difference between answering the question to be answered and the already answered question; m1 represents the number of main knowledge points in the question to be answered that match the main knowledge points in the already answered question, which can also be called the number of second main knowledge points that match the first main knowledge point; n1 represents the number of secondary knowledge points in the question to be answered that match the secondary knowledge points in the already answered question, which can also be called the number of second secondary knowledge points that match the first secondary knowledge point; m2 represents the number of main knowledge points in the question to be answered that match the secondary knowledge points in the already answered question, which can also be called the number of second main knowledge points that match the first secondary knowledge point; n2 represents the number of secondary knowledge points in the question to be answered that match the main knowledge points in the already answered question, which can also be called the number of second secondary knowledge points that match the first main knowledge point. l1 represents the number of main knowledge points that were not hit between the questions to be answered and the questions that have already been answered; λ1 represents the penalty factor for the main knowledge points that were not hit between the questions to be answered and the questions that have already been answered, which is the aforementioned first penalty coefficient; l2 represents the number of secondary knowledge points that were not hit between the questions to be answered and the questions that have already been answered; λ2 represents the penalty factor for the secondary knowledge points that were not hit between the questions to be answered and the questions that have already been answered, which is the aforementioned second penalty coefficient.

[0056] It should be noted that the above is only one implementation example, and other methods can also be used, which are not limited here. In addition, in practical applications, the above weights or coefficients can be flexibly set according to actual needs. For items that are not needed, the corresponding weights or coefficients can also be set to 0.

[0057] Understandably, if there are multiple questions already answered, the forgetting information between the questions to be answered and each already answered question can be predicted separately, and a forgetting matrix can be constructed as: R f ={r1,R2,R3.....R i For any question that has been answered, the above method can be used to determine the corresponding forgotten information.

[0058] By using the above methods, the current forgetting status of users can be more reasonably measured based on the hit results of primary and secondary knowledge points, making the obtained forgetting information more objective and accurate.

[0059] In some implementations, the first question information includes first text information and first knowledge point information; the second question information includes second text information and second knowledge point information; based on this, the above step S106, that is, the step of determining the similarity information between the answered questions and the unanswered questions based on the first question information and the second question information, can be performed with reference to the following steps a and c:

[0060] Step a: Determine the text similarity between the answered questions and the unanswered questions based on the first and second text information. In some implementation examples, the text similarity between the first and second questions can be obtained based on the similarity between the sentence vectors of the first question and the sentence vectors of the second question.

[0061] Step b: Based on the first knowledge point information and the second knowledge point information, determine the knowledge point similarity between the answered questions and the unanswered questions. In some implementation examples, the knowledge point similarity between the answered questions and the unanswered questions can be determined based on the knowledge point matching results between them. In other implementation examples, a first matching result between the first knowledge point information and multiple preset target knowledge points can be determined, and a comprehensive knowledge vector of the answered questions can be generated based on the first matching result. Similarly, a second matching result between the second knowledge point information and multiple preset target knowledge points can be determined, and a comprehensive knowledge vector of the unanswered questions can be generated based on the second matching result. Then, the knowledge point similarity between the answered questions and the unanswered questions can be determined based on the comprehensive knowledge vectors of the answered questions and the unanswered questions.

[0062] Step c: Generate a comprehensive similarity score between answered questions and unanswered questions based on text similarity and knowledge point similarity. For example, the text similarity and knowledge point similarity can be weighted and summed to obtain the comprehensive similarity score.

[0063] To further improve the accuracy and reliability of question similarity, the first question information also includes first answer results from multiple historical users for previously answered questions; the second question information also includes second answer results from multiple historical users for questions yet to be answered. Historical users are different from target users; they are users who have previously answered both previously answered questions and questions yet to be answered by the target user. In this embodiment, it is fully considered that if two questions are similar, the users' answers to these two questions will also be related. Therefore, this embodiment can statistically analyze the answer results of multiple historical users to measure the correlation between two questions from the perspective of answer results. If two questions are related, it indicates that the two questions have a certain degree of similarity. Based on this, step c can be performed with reference to steps c1 to c3 as follows:

[0064] Step c1: Based on the first and second answer result information, a correlation analysis algorithm is used to perform a correlation analysis on the answered questions and the unanswered questions to obtain the correlation analysis results. For example, by analyzing the answers of a large number of users to two questions, if most users can also answer the unanswered question correctly after answering the already answered question correctly, or answer the unanswered question incorrectly after answering the already answered question incorrectly, then the two questions can be considered to have a certain degree of similarity. Based on the above principle, a correlation analysis algorithm can be used to evaluate the relevance of the questions based on the answer result information. This embodiment of the disclosure does not limit the specific type of correlation analysis algorithm. For example, the correlation analysis algorithm can be the Phi correlation analysis method, the Spearman rank correlation coefficient method, etc., and a suitable correlation analysis algorithm can be flexibly selected according to the needs.

[0065] Step c2: Determine the similarity of answers between the already answered questions and the questions to be answered based on the correlation analysis results.

[0066] For example, the correlation analysis results are represented in the form of correlation values. The similarity of the answer results is positively correlated with the correlation values. In some specific examples, the correlation values ​​range from 0 to 1, so the correlation values ​​can be directly used as the similarity of the answer results. Taking the correlation analysis results including the phi correlation value as an example, the similarity of the answer results is positively correlated with the phi correlation value. In some specific examples, the phi correlation value can be directly used as the similarity of the answer results.

[0067] Step c3: Based on text similarity, knowledge point similarity, and answer result similarity, generate a comprehensive similarity between the answered questions and the unanswered questions.

[0068] In some implementation examples, a comprehensive similarity score can be obtained by weighted summation of text similarity, knowledge point similarity, and answer result similarity. This comprehensive similarity score can fully encompass the similarity assessment results across all dimensions and match answered and unanswered questions across multiple dimensions. It should be noted that this embodiment does not calculate the similarity of two questions by uniformly combining multiple pieces of information. Instead, it calculates the question similarity for each dimension (text, knowledge point, answer result dimension) separately. This method allows for targeted measurement of similarity within each dimension, resulting in a more accurate and reliable comprehensive similarity score.

[0069] If there are multiple questions already answered and Qn questions remain to be answered, then the question similarity information matrix R can also be obtained using the above method. s ={A n1 A n2 A n3 ,……A ni}, where Ani Let n be the question similarity between the n questions to be answered and the i-th question that has already been answered.

[0070] In some implementations, step S108, which is the step of predicting the target user's accuracy in answering the questions based on forgotten information and similarity information, can be performed with reference to steps one through three below:

[0071] Step one involves fusing the forgotten information and similarity information to obtain fused information. In practical applications, the forgotten information and similarity information can be represented by specific numerical values ​​or vector forms; there are no specific restrictions.

[0072] In some specific implementation examples, normalization can be performed first on the forgotten information and the similar information respectively; then, the normalized forgotten information and the normalized similar information can be summed to obtain the fused information.

[0073] Assume the forgotten information is R f Similar information is R s Then the fused information can be: R = softmaxR f +softmaxR s Taking the example that forgotten information and similar information are specific numerical values, such as forgetting information being represented by forgetting values ​​and similar information being represented by similarity values, in order to reasonably integrate the two, we can first normalize both to the same scale before integrating them.

[0074] Step two: Obtain the target user's answer results for the already answered questions. The target user's answer results for the already answered questions are the target answer result information. The answer result information can be represented by numerical values ​​such as 0 or 1, where 0 indicates an incorrect answer and 1 indicates a correct answer.

[0075] Step 3: Based on the fused information and the target user's answer results, predict the target user's accuracy rate for the questions to be answered. The fused information can effectively present the similarity between the answered questions and the questions to be answered, as well as the user's forgetting at the knowledge point level. Combined with the target user's answer results for the answered questions, the accuracy rate of the target user's answer for the questions to be answered can be comprehensively predicted.

[0076] For example, the first question information includes the first text information; step three can be implemented by referring to the following steps (1) and (2):

[0077] Step (1): Based on the target answer result information and the first text information, obtain the interaction vector of the answered question. This interaction vector can present both the information of the question text and the user's answer result, that is, it can be understood as presenting the situation of the user's interaction with the question.

[0078] In some specific implementation examples, steps 1 to 3 can be performed as follows:

[0079] Step 1: Based on the first text information, generate sentence vectors for the answered questions. For example, word embedding can be performed on the first text information of the answered questions to obtain word vectors corresponding to the answered questions; then, the Smooth Inverse Frequency (SIF) algorithm can be used to process the word vectors corresponding to the answered questions to obtain sentence vectors corresponding to the answered questions.

[0080] Step 2: Perform dimensional expansion processing based on the target response information to obtain a response result vector with the same dimension as the sentence vector. Assume the target response information is r. i r i The value can be either 0 or 1. To facilitate subsequent processing in conjunction with sentence vectors, it can be extended to the dimension d of the sentence vector, denoted as r. i =[r i ,r i ,r i ...r i ].

[0081] Step 3: Generate the interaction vector for the answered questions based on the sentence vector and the answer result vector. For example, the sentence vector and the answer result vector can be merged to obtain the interaction vector for the answered questions. In specific implementation, step 3 can be performed as follows: Steps 3.1 to 3.3

[0082] Step 3.1: Perform Boolean operations on the sentence vector and the answer result vector to obtain the vector after Boolean operations. Boolean operations can enhance the influence of the answer result on the sentence vector.

[0083] Step 3.2: Perform word segmentation processing on the first text information, and obtain the word position vectors of the first text information based on the word segmentation results. That is, perform word segmentation processing on the first text information to obtain word vectors. Since each word has a certain order in the text, the word position vectors can be obtained based on the position information of each word vector.

[0084] Step 3.3: Summate the vector after Boolean operation with the word position vector to obtain the interaction vector of the answered questions.

[0085] Based on steps 3.1 to 3.3 above, the interaction vectors of the answered questions can be obtained. It should be noted that the above method not only combines sentence vectors and answer result vectors, but also introduces word position vectors, making the resulting interaction vectors more reasonable.

[0086] Step (2) involves using a pre-defined self-attention model to predict the target user's accuracy in answering questions based on fused information and interaction vectors. The self-attention model predicts accuracy based on a self-attention mechanism. The essence of the attention mechanism is a mechanism that uses a set of weighted coefficients learned autonomously by the model and dynamically weights specific data to improve the overall performance of the model. In practical applications, the initial self-attention model can be trained using question samples labeled with the answers, resulting in a well-trained model that can be directly applied to the scenario of predicting accuracy.

[0087] To facilitate understanding, the following is a specific implementation example of using a pre-defined self-attention model to predict the accuracy of a target user's answers to a question:

[0088] Obtain the sentence vector E of the question to be answered. n And the dimension of the sentence vector is d, and the interaction vector of the i-th answered question is obtained. And obtain the query parameter W of the self-attention model. Q Index parameter W K Content parameter W V Based on this, the first parameter t can be obtained. i Specifically:

[0089] The answered questions can then be normalized to obtain the second parameter α. i ,in, Based on this, the third parameter β can be further obtained. i =λα i +(1-λ)R i Where λ is a preset value, and R i The fused information for the i-th answered question can be obtained using the aforementioned method. Furthermore, the fourth parameter can be calculated:

[0090] Based on the above, by performing excitation processing through a feedforward neural network, the value can be obtained.

[0091] F = ReLu(oW) 1 +b 1 W 2 +b 2

[0092] Where ReLU is the activation function, W 1 and W 2 All are model weight parameters, b 1 and b 2 These are all model weight parameters, specifically the intercept.

[0093] Finally, the accuracy p of answering the questions can be obtained through a logistic regression layer.

[0094]

[0095] By using the above method, the self-attention mechanism can focus more on the important features in the fused information and interaction vectors, resulting in more accurate and reliable prediction results.

[0096] Building upon the foregoing, for ease of understanding, this disclosure also provides a specific implementation example of a method for predicting answer accuracy, see [link to relevant documentation]. Figure 2 The flowchart shown illustrates a method for predicting answer accuracy. It shows that forgotten information and similarity information need to be calculated separately for each question to be answered. Forgotten information is determined based on main knowledge points and secondary knowledge points, while similarity information is determined based on a combination of text similarity, relevance of answer results, and similarity of knowledge points. Subsequently, the forgotten information and similarity information are processed using a self-attention model to predict the answer accuracy of the question to be answered. Figure 2 This is just an example. In practical applications, it may not be necessary to further divide the main and secondary knowledge points, or only one or two of the following can be performed: text similarity calculation, answer result relevance calculation, and knowledge point similarity calculation. The specific settings can be flexible.

[0097] In summary, the above-mentioned method for predicting answer accuracy fully considers the phenomenon of user forgetting and the impact of question similarity on answer accuracy. Therefore, it can obtain forgotten information and similarity information between answered questions and unanswered questions, and combine forgotten information and similarity information to predict the user's answer accuracy. Moreover, the method of obtaining forgotten information and similarity information is more accurate and reliable, which can further make the prediction results more consistent with the actual situation and help to further improve the accuracy of the prediction results.

[0098] Based on the foregoing, this disclosure further provides a method for recommending topics. Figure 3 This is a flowchart illustrating a question recommendation method provided in an embodiment of this disclosure. This method can be executed by a question recommendation device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. Figure 3 As shown, the method mainly includes the following steps S302 to S306:

[0099] Step S302: Obtain multiple questions to be predicted. These questions are all those that the user has not yet answered.

[0100] Step S304: Obtain the target user's answer accuracy rate for each question to be predicted; wherein, the question to be predicted is taken as the aforementioned question to be answered, and the aforementioned answer accuracy rate prediction method is used to determine the target user's answer accuracy rate for each question to be predicted.

[0101] Step S306: Recommend target questions to the target user based on the accuracy rate of each question to be predicted.

[0102] In this embodiment of the disclosure, multiple questions to be predicted can be obtained first, and questions can be recommended based on the accuracy of the answers to these questions. For example, the user's knowledge level can be better assessed based on the accuracy of the user's answers to each question to be predicted, and questions can be recommended to the user based on this, which can effectively ensure the rationality of the target questions recommended to the user.

[0103] In some implementation examples, step S306 above, which is the step of recommending target questions to the target user based on the accuracy rate of each question to be predicted, can be performed with reference to steps S1 to S3 as follows:

[0104] Step S1: For each question to be predicted, determine the difficulty range to which the accuracy of answering the question belongs from a set of multiple difficulty ranges; where each difficulty range corresponds to a difficulty level.

[0105] In other words, multiple difficulty levels can be pre-set, each corresponding to a difficulty range. The difficulty range can be a range of accuracy rates. For example, it can be graded as follows: [0-0.2) corresponds to "Difficult"; [0.2-0.4) corresponds to "Relatively Difficult"; [0.4-0.6) corresponds to "Medium"; [0.6-0.8) corresponds to "Relatively Easy"; [0.8,1) corresponds to "Easy". For instance, assuming the target user's accuracy rate for the nth question to be predicted is 0.7, then the difficulty range for the nth question to be predicted relative to the target user is "Relatively Easy".

[0106] Step S2: Determine the target difficulty level based on the statistical distribution of the difficulty ranges of the multiple questions to be predicted. Specifically, the target difficulty level can be determined based on the statistical distribution of the difficulty ranges of the multiple questions to be predicted, and the theoretical difficulty level marked for each question. Assuming there are 10 questions to be predicted, for the target user, 8 questions are "difficult," 1 is "easy," and 1 is "medium." However, the theoretical difficulty level (i.e., the difficulty level assessed by professionals) of the 10 questions in the question bank is actually "medium." This indicates that the target user's current ability level is not high, and a typical medium-difficulty question would be difficult for the target user. Therefore, the target difficulty level can be determined to include "relatively easy" and / or "easy."

[0107] Step S3: Recommend target questions to the target user based on the target difficulty level. For example, target questions can be selected from a candidate question set and then provided to the target user. The candidate question set includes questions to be predicted and / or questions from a preset question bank, and the target questions include questions labeled with the target difficulty level. This embodiment of the disclosure does not limit the number of target questions.

[0108] By using the above methods, users' abilities can be reasonably assessed, and questions that match their abilities can be recommended. This not only ensures a good user experience but also effectively improves users' learning efficiency.

[0109] Corresponding to the aforementioned method for predicting answer accuracy, this disclosure also provides a device for predicting answer accuracy. Figure 4 This is a schematic diagram of a device for predicting the accuracy of responses, provided in an embodiment of this disclosure. This device can be implemented by software and / or hardware and is generally integrated into an electronic device. Figure 4 As shown, the accuracy prediction device 400 includes:

[0110] The information acquisition module 402 is used to acquire the first question information and first time information corresponding to the questions already answered by the target user, as well as the second question information and second time information corresponding to the questions to be answered.

[0111] The forgotten information determination module 404 is used to determine the forgotten information between the answered questions and the unanswered questions based on the first question information, the first time information, the second question information, and the second time information.

[0112] The similarity information determination module 406 is used to determine the similarity information between the answered questions and the unanswered questions based on the first question information and the second question information;

[0113] The accuracy prediction module 408 is used to predict the accuracy of a target user's answers to questions based on forgotten information and similar information.

[0114] The above-mentioned accuracy prediction device for answering questions provided in this embodiment of the present disclosure fully considers the user's forgetting phenomenon and the impact of question similarity on the accuracy of answering questions. Therefore, it can obtain forgetting information and similarity information between the already answered questions and the questions to be answered, and combine the forgetting information and similarity information to predict the user's accuracy of answering questions, so that the prediction results are more in line with the actual situation and help to further improve the accuracy of the prediction results.

[0115] In some implementations, the first question information includes first main knowledge point information and first secondary knowledge point information, and the second question information includes second main knowledge point information and second secondary knowledge point information;

[0116] The forgotten information determination module 404 is specifically used to: determine the knowledge point hit result of the question to be answered relative to the already answered question based on the first main knowledge point information, the first secondary knowledge point information, the second main knowledge point information, and the second secondary knowledge point information; and determine the forgotten information between the already answered question and the question to be answered based on the knowledge point hit result, the first time information, and the second time information.

[0117] In some implementations, the knowledge point matching results include: the number of times the second primary knowledge point matches the first primary knowledge point, the number of times the second primary knowledge point matches the first secondary knowledge point, the number of times the second secondary knowledge point matches the first primary knowledge point, and the number of times the second secondary knowledge point matches the first secondary knowledge point.

[0118] In some embodiments, the forgetting information determination module 404 is specifically used to: determine the time difference information between the answered questions and the unanswered questions based on the first time information and the second time information; obtain a preset plurality of forgetting weights; wherein the plurality of forgetting weights include: a first forgetting weight corresponding to the second main knowledge point hitting the first main knowledge point, a second forgetting weight corresponding to the second main knowledge point hitting the first secondary knowledge point, a third forgetting weight corresponding to the second secondary knowledge point hitting the first main knowledge point, and a fourth forgetting weight corresponding to the second secondary knowledge point hitting the first secondary knowledge point; and determine the forgetting information between the answered questions and the unanswered questions based on the time difference information, the plurality of forgetting weights, and the knowledge point hitting result.

[0119] In some implementations, the forgetting information determination module 404 is specifically used to: obtain a first penalty item and a second penalty item; wherein, the first penalty item is determined based on the number of times the first main knowledge point is not hit by the second main knowledge point and the second secondary knowledge point and a preset first penalty coefficient; the second penalty item is determined based on the number of times the first secondary knowledge point is not hit by the second main knowledge point and the second secondary knowledge point and a preset second penalty coefficient; perform weighted averaging processing based on the time difference information, the multiple forgetting weights and the knowledge point hit results to determine a comprehensive forgetting item; and determine the forgetting information between the answered questions and the unanswered questions based on the comprehensive forgetting item, the first penalty item and the second penalty item.

[0120] In some implementations, the first question information includes first text information and first knowledge point information; the second question information includes second text information and second knowledge point information; the similarity information determination module 406 is specifically used to: determine the text similarity between the answered question and the unanswered question based on the first text information and the second text information; determine the knowledge point similarity between the answered question and the unanswered question based on the first knowledge point information and the second knowledge point information; and generate a comprehensive similarity between the answered question and the unanswered question based on the text similarity and the knowledge point similarity.

[0121] In some implementations, the first question information further includes first answer result information of multiple historical users answering the already answered questions; the second question information further includes second answer result information of the multiple historical users answering the unanswered questions; the similarity information determination module 406 is specifically used to: perform correlation analysis on the already answered questions and the unanswered questions based on the first answer result information and the second answer result information, and obtain correlation analysis results; determine the similarity of answer results between the already answered questions and the unanswered questions based on the correlation analysis results; and generate a comprehensive similarity between the already answered questions and the unanswered questions based on the text similarity, the knowledge point similarity, and the answer result similarity.

[0122] In some implementations, the accuracy prediction module 408 is specifically used to: perform fusion processing on the forgotten information and the similarity information to obtain fused information; obtain the target answer result information of the target user for the already answered questions; and predict the accuracy of the target user's answer for the unanswered questions based on the fused information and the target answer result information.

[0123] In some implementations, the accuracy prediction module 408 is specifically used to: perform normalization processing on the forgotten information and the similarity information respectively; and perform summation processing on the normalized forgotten information and the normalized similarity information to obtain fused information.

[0124] In some implementations, the first question information includes first text information; the accuracy prediction module 408 is specifically used to: obtain the interaction vector of the answered question based on the target answer result information and the first text information; and predict the target user's answer accuracy for the question to be answered using a preset self-attention model based on the fused information and the interaction vector.

[0125] In some implementations, the accuracy prediction module 408 is specifically used to: generate a sentence vector of the answered questions based on the first text information; perform dimensional expansion processing based on the target answer result information to obtain an answer result vector with the same dimension as the sentence vector; and generate an interaction vector of the answered questions based on the sentence vector and the answer result vector.

[0126] In some implementations, the accuracy prediction module 408 is specifically used to: perform Boolean operations on the sentence vector and the answer result vector to obtain a Boolean-operated vector; perform word segmentation processing based on the first text information, and obtain the word position vector of the first text information according to the word segmentation result; and sum the Boolean-operated vector and the word position vector to obtain the interaction vector of the answered question.

[0127] The accuracy prediction device provided in this disclosure can execute the accuracy prediction method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the execution method.

[0128] Corresponding to the aforementioned question recommendation method, this disclosure also provides a question recommendation device. Figure 5 This is a schematic diagram of a topic recommendation device provided in an embodiment of this disclosure. The device can be implemented by software and / or hardware and is generally integrated into an electronic device. Figure 5 As shown, the recommended device 500 includes:

[0129] The question acquisition module 502 is used to acquire multiple questions to be predicted.

[0130] The accuracy acquisition module 504 is used to acquire the accuracy rate of the target user's answers to each question to be predicted; wherein, the accuracy rate is determined based on the aforementioned prediction method for accuracy rate.

[0131] The question recommendation module 506 is used to recommend target questions to target users based on the accuracy rate of each question to be predicted.

[0132] The question recommendation device provided in this embodiment can use the aforementioned method for predicting the accuracy of a user's answers to multiple questions to make question recommendations, thereby effectively ensuring the rationality of the target questions recommended to the user.

[0133] In some implementations, the question recommendation module 506 is specifically used to: for each question to be predicted, determine the difficulty range to which the accuracy of answering the question to be predicted belongs from a plurality of pre-set difficulty ranges; wherein, each difficulty range corresponds to a difficulty level; determine a target difficulty level based on the distribution statistics of the difficulty ranges to which the plurality of questions to be predicted belong; and recommend target questions to the target user according to the target difficulty level.

[0134] In some implementations, the question recommendation module 506 is specifically used to: select a target question from a set of candidate questions; wherein the set of candidate questions includes the question to be predicted and / or questions in a preset question bank, and the target question includes questions marked with a target difficulty level; and provide the target question to the target user.

[0135] The question recommendation device provided in this disclosure can execute the question recommendation method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects of the method execution.

[0136] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device embodiments can be referred to the corresponding process in the method embodiments, and will not be repeated here.

[0137] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0138] Exemplary embodiments of this disclosure also provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the electronic device to perform a method according to an embodiment of this disclosure.

[0139] Exemplary embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to embodiments of this disclosure.

[0140] Exemplary embodiments of this disclosure also provide a computer program product, including a computer program, wherein, when executed by a processor of a computer, the computer program is used to cause the computer to perform a method according to an embodiment of this disclosure.

[0141] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this disclosure. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0142] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the methods provided in embodiments of this disclosure. The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0143] refer to Figure 6 The present invention describes a structural block diagram of an electronic device 600 that can serve as a server or client of the present disclosure, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0144] like Figure 6As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0145] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, output unit 607, storage unit 608, and communication unit 609. Input unit 606 can be any type of device capable of inputting information to electronic device 600. Input unit 606 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 607 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 608 may include, but is not limited to, disks and optical discs. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0146] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above. For example, in some embodiments, the method for predicting answer accuracy or the method for recommending questions can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. In some embodiments, the computing unit 601 can be configured by any other suitable means (e.g., by means of firmware) to perform the method for predicting answer accuracy or the method for recommending questions.

[0147] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0148] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0149] As used in this disclosure, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0150] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0151] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0152] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0153] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0154] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the accuracy of responses, comprising: Obtain the first question information and first time information corresponding to the questions already answered by the target user, as well as the second question information and second time information corresponding to the questions to be answered; Based on the first question information, the first time information, the second question information, and the second time information, the forgetting information between the answered questions and the unanswered questions is determined. The forgetting information is determined by weighted summation based on multiple forgetting weights and the number of knowledge points hits corresponding to each forgetting weight, combined with time difference information, using a preset forgetting algorithm. Based on the first question information and the second question information, determine the similarity information between the answered questions and the unanswered questions; Based on the forgotten information and the similarity information, predict the accuracy rate of the target user's answers to the questions to be answered; wherein the steps include: The forgotten information and the similarity information are fused together to obtain fused information; Obtain the target user's answer result information for the already answered questions; Based on the fused information and the target answer result information, the accuracy rate of the target user's answer to the question to be answered is predicted.

2. The method for predicting answer accuracy as described in claim 1, wherein, The first question information includes the first main knowledge point information and the first secondary knowledge point information; the second question information includes the second main knowledge point information and the second secondary knowledge point information. The step of determining the forgotten information between the answered questions and the unanswered questions based on the first question information, the first time information, the second question information, and the second time information includes: Based on the first primary knowledge point information, the first secondary knowledge point information, the second primary knowledge point information, and the second secondary knowledge point information, determine the knowledge point hit result of the question to be answered relative to the already answered question; Based on the knowledge point hit results, the first time information, and the second time information, the forgetting information between the answered questions and the unanswered questions is determined.

3. The method for predicting answer accuracy as described in claim 2, wherein, The knowledge point hit results include: the number of times the second primary knowledge point hits the first primary knowledge point, the number of times the second primary knowledge point hits the first secondary knowledge point, the number of times the second secondary knowledge point hits the first primary knowledge point, and the number of times the second secondary knowledge point hits the first secondary knowledge point.

4. The method for predicting answer accuracy as described in claim 2, wherein, The step of determining the forgetting information between the answered questions and the unanswered questions based on the knowledge point hit result, the first time information, and the second time information includes: Based on the first time information and the second time information, determine the time difference information between the answered questions and the unanswered questions; Obtain multiple preset forgetting weights; wherein, the multiple forgetting weights include: a first forgetting weight corresponding to the second main knowledge point hitting the first main knowledge point, a second forgetting weight corresponding to the second main knowledge point hitting the first secondary knowledge point, a third forgetting weight corresponding to the second secondary knowledge point hitting the first main knowledge point, and a fourth forgetting weight corresponding to the second secondary knowledge point hitting the first secondary knowledge point. Based on the time difference information, the multiple forgetting weights, and the knowledge point hit results, the forgetting information between the answered questions and the unanswered questions is determined.

5. The method for predicting answer accuracy as described in claim 4, wherein, The step of determining the forgetting information between the answered questions and the unanswered questions based on the time difference information, the multiple forgetting weights, and the knowledge point hit results includes: Obtain a first penalty item and a second penalty item; wherein, the first penalty item is determined based on the number of times the first primary knowledge point is not hit by the second primary knowledge point and the second secondary knowledge point and a preset first penalty coefficient; the second penalty item is determined based on the number of times the first secondary knowledge point is not hit by the second primary knowledge point and the second secondary knowledge point and a preset second penalty coefficient; The weighted average is calculated based on the multiple forgetting weights and the number of knowledge point hits of the questions to be answered relative to the questions already answered. The weighted average result is then multiplied by the time difference information to obtain the result value. The negative value of the result value is then used as the exponent of the natural constant e to obtain the comprehensive forgetting term. Based on the comprehensive forgetting item, the first penalty item, and the second penalty item, forgetting information between the answered questions and the unanswered questions is determined.

6. The method for predicting answer accuracy as described in claim 1, wherein, The first question information includes first text information and first knowledge point information; the second question information includes second text information and second knowledge point information. The step of determining the similarity information between the answered questions and the unanswered questions based on the first question information and the second question information includes: Based on the first text information and the second text information, determine the text similarity between the answered questions and the unanswered questions; Based on the first knowledge point information and the second knowledge point information, determine the knowledge point similarity between the answered questions and the unanswered questions; Based on the text similarity and the knowledge point similarity, a comprehensive similarity is generated between the answered questions and the unanswered questions.

7. The method for predicting answer accuracy as described in claim 6, wherein, The first question information also includes first answer result information of multiple historical users answering the already answered questions; the second question information also includes second answer result information of the multiple historical users answering the unanswered questions. The step of generating a comprehensive similarity between the answered questions and the unanswered questions based on the text similarity and the knowledge point similarity includes: Based on the first answer result information and the second answer result information, a correlation analysis algorithm is used to perform a correlation analysis on the already answered questions and the questions to be answered, and the correlation analysis results are obtained. The similarity of the answers between the already answered questions and the questions to be answered is determined based on the correlation analysis results. Based on the text similarity, the knowledge point similarity, and the answer result similarity, a comprehensive similarity is generated between the answered questions and the unanswered questions.

8. The method for predicting answer accuracy as described in claim 1, wherein, The step of fusing the forgotten information and the similarity information to obtain fused information includes: Normalization is performed based on the forgotten information and the similarity information, respectively. The forgotten information and the similarity information after normalization are summed to obtain fused information.

9. The method for predicting the accuracy of responses as described in claim 1, wherein, The first question information includes first text information; the step of predicting the accuracy rate of the target user's answer to the question to be answered based on the fused information and the target answer result information includes: Based on the target answer result information and the first text information, the interaction vector of the answered question is obtained. The interaction vector of the answered question is used to present the information of the question text and the target user's answer result. Based on the fused information and the interaction vector, a preset self-attention model is used to predict the accuracy of the target user's answers to the questions to be answered; The step of obtaining the interaction vector of the answered question based on the target answer result information and the first text information includes: Based on the first text information, generate sentence vectors for the answered questions; Based on the target response information, dimensional expansion processing is performed to obtain a response result vector with the same dimension as the sentence vector; Based on the sentence vector and the answer result vector, generate the interaction vector of the answered question.

10. The method for predicting answer accuracy as described in claim 9, wherein, The step of generating the interaction vector of the answered question based on the sentence vector and the answer result vector includes: Perform a Boolean operation on the sentence vector and the answer result vector to obtain the vector after the Boolean operation; Based on the first text information, word segmentation processing is performed, and the word position vectors of the first text information are obtained according to the word segmentation results; The vector obtained after the Boolean operation is summed with the word position vector to obtain the interaction vector of the answered question.

11. A question recommendation method, including: Get multiple questions to be predicted; Obtain the accuracy rate of the target user's answers to each of the predicted questions; wherein the accuracy rate is determined based on the prediction method for accuracy rate according to any one of claims 1 to 10; Based on the accuracy rate of answering each of the predicted questions, target questions are recommended to the target user.

12. The question recommendation method as described in claim 11, wherein, The step of recommending target questions to the target user based on the accuracy rate of answering each of the predicted questions includes: For each question to be predicted, the difficulty range to which the accuracy of answering the question belongs is determined from a set of multiple difficulty ranges; where each difficulty range corresponds to a difficulty level. Based on the statistical distribution of the difficulty ranges to which the multiple questions to be predicted belong, the target difficulty level is determined; Based on the target difficulty level, target questions are recommended to the target user.

13. The question recommendation method as described in claim 12, wherein, The step of recommending target questions to the target user based on the target difficulty level includes: Select a target question from the candidate question set; wherein the candidate question set includes the question to be predicted and / or questions in a preset question bank, and the target question includes questions marked with a target difficulty level; The target question is provided to the target user.

14. A device for predicting the accuracy of responses, comprising: The information acquisition module is used to acquire the first question information and first time information corresponding to the questions already answered by the target user, as well as the second question information and second time information corresponding to the questions to be answered. The forgetting information determination module is used to determine the forgetting information between the answered questions and the unanswered questions based on the first question information, the first time information, the second question information, and the second time information. The forgetting information is determined by weighted summation based on multiple forgetting weights and the number of knowledge points hits corresponding to each forgetting weight, combined with time difference information, and using a preset forgetting algorithm. The similarity information determination module is used to determine the similarity information between the answered questions and the unanswered questions based on the first question information and the second question information; The accuracy prediction module is used to predict the accuracy rate of the target user's answers to the questions to be answered, based on the forgotten information and the similarity information; wherein the steps include: The forgotten information and the similarity information are fused together to obtain fused information; Obtain the target user's answer result information for the already answered questions; Based on the fused information and the target answer result information, the accuracy rate of the target user's answer to the question to be answered is predicted.

15. A question recommendation device, comprising: The question acquisition module is used to acquire multiple questions to be predicted; An accuracy acquisition module is used to acquire the accuracy rate of a target user's answers to each of the questions to be predicted; wherein the accuracy rate is determined based on the prediction method for accuracy rate according to any one of claims 1 to 10; The question recommendation module is used to recommend target questions to the target user based on the accuracy rate of answering each question to be predicted.

Images