Knowledge practice method and device, computer equipment and storage medium

Abstract

The invention discloses a knowledge practice method and device, computer equipment and a storage medium, and the method comprises the steps: extracting a plurality of knowledge questions from a knowledge question bank matched with a knowledge practice theme, obtaining a first knowledge question set, and guaranteeing that the knowledge questions for knowledge practice are knowledge questions related to the practice theme. The knowledge problem of knowledge practice is targeted; receiving an initial response answer set of the first knowledge question set, determining practice result data of a round of knowledge practice based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set, and re-extracting questions from the knowledge question bank based on the practice result data, according to the technical scheme, the knowledge problems are dynamically and circularly extracted in combination with the practice condition, it is ensured that the re-extracted knowledge problem set can accurately reflect the knowledge practice condition, and multiple rounds of practice are conducted on the dynamically extracted knowledge problem set, so that knowledge practice is more targeted, and the knowledge practice effect is greatly improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a knowledge practice method, apparatus, computer equipment and storage medium. Background Art

[0002] With the development of society, enterprises need to continuously upgrade their employee training needs. The traditional manual collective training method is inefficient and consumes a lot of manpower and time. Therefore, knowledge practice systems have emerged, such as the training system, which is used to simulate real customer service scenarios in customer service scenarios, providing students with an efficient and convenient learning platform, thereby improving the company's service level and customer service satisfaction.

[0003] The current tutoring system manually extracts key knowledge points from knowledge materials to form a question bank, and attaches reference answers. During the student practice stage, questions are extracted from the question bank for students to practice, and scores are given based on the students' answers. The students' mastery of the knowledge points is assessed based on the scores. Summary of the invention

[0004] In response to the above technical problems, embodiments of the present application provide a knowledge practice method, apparatus, computer device and storage medium.

[0005] In a first aspect, an embodiment of the present application provides a knowledge practice method, comprising:

[0006] In response to the knowledge practice instruction, obtaining a knowledge practice topic from the knowledge practice instruction;

[0007] Extracting knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer;

[0008] receiving an initial response answer set of the first knowledge question set, and determining exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set;

[0009] Based on the practice result data, questions are re-drawn from the knowledge question bank to obtain a second knowledge question set, and the next round of knowledge practice is performed on the second knowledge question set until the practice end condition is met.

[0010] In a second aspect, an embodiment of the present application provides a knowledge training device, including:

[0011] An acquisition module, configured to acquire a knowledge practice topic from the knowledge practice instruction in response to the knowledge practice instruction;

[0012] A question extraction module is used to extract knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer;

[0013] a determination module, configured to receive an initial response answer set of the first knowledge question set, and determine exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set;

[0014] The practice module is used to re-select questions from the knowledge question bank based on the practice result data to obtain a second knowledge question set, and perform the next round of knowledge practice on the second knowledge question set until the practice end condition is met.

[0015] In a third aspect, an embodiment of the present application further provides a computer device, including a memory storing a plurality of instructions; a processor loads instructions from the memory to execute the steps of any one of the knowledge practice methods provided in the embodiments of the present application.

[0016] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores a plurality of instructions suitable for loading by a processor to execute the steps of any one of the knowledge practice methods provided in the embodiments of the present application.

[0017] In a fifth aspect, an embodiment of the present application further provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements the steps of any one of the knowledge practice methods provided in the embodiments of the present application.

[0018] By adopting the scheme of the application embodiment, knowledge questions are extracted from the knowledge question bank that matches the knowledge practice theme to obtain a first knowledge question set, which can ensure that the knowledge questions for knowledge practice are knowledge questions related to the practice theme, so that the knowledge questions for knowledge practice are targeted; then, an initial response answer set of the first knowledge question set is received, and based on the initial response answer set and the reference answers corresponding to each knowledge question in the first knowledge question set, the practice result data of a round of knowledge practice is determined, and based on the practice result data, questions are re-extracted from the knowledge question bank to obtain a second knowledge question set, thereby realizing dynamic and cyclic extraction of knowledge questions combined with the practice situation, realizing real-time dynamic question extraction, ensuring that the re-extracted knowledge question set can accurately reflect the knowledge practice situation, and performing multiple rounds of practice on the dynamically extracted knowledge question set, thereby making the knowledge practice more targeted and greatly improving the knowledge practice effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 is a schematic diagram of an application environment of the knowledge practice method provided in an embodiment of the present application;

[0021] Figure 2 It is a schematic diagram of a knowledge practice process provided by the relevant technology;

[0022] Figure 3 It is a schematic diagram of an embodiment of the knowledge practice method provided in the embodiments of the present application;

[0023] Figure 4 It is a flowchart of another embodiment of the knowledge practice method provided in the embodiments of the present application;

[0024] Figure 5 is a schematic diagram of the target prompt word reasoning process provided in the embodiment of the present application;

[0025] Figure 6 is another schematic diagram of the target prompt word reasoning process provided in the embodiments of the present application;

[0026] Figure 7 It is a schematic diagram of generating text knowledge data provided in an embodiment of the present application;

[0027] Figure 8 This is a schematic diagram of the knowledge practice process in the voice mode provided in the embodiment of the present application.

[0028] Fig. 9 is a schematic diagram of the structure of the knowledge practice device provided in the embodiment of the present application;

[0029] Fig.10 It is a schematic diagram of the internal structure of the computer device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application. At the same time, in the description of the embodiments of the present application, the terms "first", "second", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more features. In the description of the embodiments of the present application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0031] In one embodiment of the present application, the knowledge practice method can be run on a local terminal device or a server. When the knowledge practice method is run on a server, the method can be implemented and executed based on a cloud interaction system, wherein the cloud interaction system includes a server and a client device.

[0032] In order to better understand the knowledge practice method, device, electronic device and storage medium provided in the embodiments of the present application, the application environment applicable to the embodiments of the present application is first described below.

[0033] See also Figure 1 , Figure 1 FIG. 1 is a schematic diagram showing an application environment of the knowledge practice method provided by an embodiment of the present application. As an implementation mode, the knowledge practice method provided by the embodiment of the present application can be applied to an electronic device. The electronic device can be as follows: Figure 1 The server 110 shown in the figure can be connected to the terminal device 120 through a network. The network is used to provide a medium for a communication link between the server 110 and the terminal device 120. The network can include various connection types, such as a wired communication link, a wireless communication link, etc., which are not limited in the embodiments of the present application. Optionally, in other embodiments, the electronic device can also be a smart phone, a laptop computer, etc.

[0034] It should be understood that Figure 1 The server 110, network and terminal device 120 are merely illustrative. Depending on the implementation requirements, there may be any number of servers, networks and terminal devices. For example, the server 110 may be a physical server or a server cluster composed of multiple servers, and the terminal device 120 may be a mobile phone, a tablet, a desktop computer, a laptop computer, and the like. It is understood that the embodiments of the present application may also allow multiple terminal devices 120 to access the server 110 at the same time.

[0035] The following is a detailed description in conjunction with the accompanying drawings. In this embodiment, the execution subject is a server as an example. It should be noted that the description order of the following embodiments is not intended to limit the preferred order of the embodiments. Although the logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in an order different from that shown in the accompanying drawings.

[0036] The knowledge practice method of this embodiment can be applied to a variety of scenarios that require automatic training, including but not limited to intelligent training scenarios in which customer service personnel receive skill training and job assessment through an intelligent training system.

[0037] In the related art, Figure 2 A schematic diagram of the knowledge practice process provided for related technologies. Figure 2 As shown, the knowledge practice system administrator extracts key points from the knowledge materials to form a question bank, attaches reference answers, and imports them into the training system. The students select specific practice scenarios to extract questions and answer the questions. The knowledge practice system scores the answers and assesses the students' mastery of the knowledge points based on the scores.

[0038] However, in the above-mentioned knowledge practice method, questions are drawn from a question bank based on specific rules during the student practice stage. It cannot be dynamically adjusted according to the student's mastery of knowledge points, making it difficult to accurately practice according to the student's personalized needs, thereby reducing the practice effect.

[0039] In order to solve the above problems, the present application provides a knowledge practice method, which can dynamically update the initial knowledge questions in real time based on the client's practice situation, ensure that the updated knowledge question set can accurately reflect the client's practice situation, and perform multiple rounds of knowledge practice on the updated knowledge question set, so that the client's knowledge practice is more targeted, which is conducive to improving the client's practice effect.

[0040] Please refer to Figure 3 The specific process of the knowledge training method can be as follows: Step 201 to Step 204, wherein:

[0041] Step 201 , in response to a knowledge practice instruction, obtaining a knowledge practice topic from the knowledge practice instruction.

[0042] The knowledge practice instruction is triggered by the user through the client to initiate knowledge practice. The knowledge practice topic refers to the knowledge topic that the client needs to practice. Taking the scenario of a company's customer service learning assessment as an example, the practice topic can be a policy and regulation topic, a speech topic, a product topic, etc., and each practice topic is configured with knowledge questions of multiple categories of knowledge points.

[0043] The knowledge practice method in this embodiment supports knowledge practice of at least one knowledge practice theme. Accordingly, the practice theme is obtained from the knowledge practice instruction, so that subsequent knowledge practice under the knowledge practice theme can be carried out in a targeted manner.

[0044] Step 202 , extracting knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer.

[0045] Among them, the knowledge question base refers to a collection of multiple knowledge questions and reference answers associated with the knowledge practice topic pre-configured on the server. It is a database used for the client to practice knowledge questions. The knowledge base can be constructed by analyzing the knowledge data associated with the knowledge practice topic through a generative artificial intelligence (Artificial Intelligence Generated Content, AIGC) model.

[0046] The first knowledge question set is a collection of multiple knowledge questions randomly extracted from the knowledge question bank, that is, a combination of multiple knowledge questions that the client practices in this round of knowledge practice, and the number of knowledge questions in the first knowledge question set is smaller than the number of knowledge questions in the knowledge question bank, that is, during a round of knowledge practice, some knowledge questions are randomly extracted for knowledge practice, which facilitates the subsequent targeted extraction of the next round of knowledge questions based on the knowledge practice of these knowledge questions, thereby improving the effectiveness of the client's knowledge practice.

[0047] Step 203, receiving an initial response answer set of the first knowledge question set, and determining exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set.

[0048] The initial response answer set is a set of responses to the knowledge questions in the first knowledge question set fed back by the client during the knowledge practice process.

[0049] The exercise result data is result data reflecting the user's mastery of the knowledge points of the first knowledge question set in a round of knowledge practice. For example, the exercise result data may be exercise score, accuracy rate, error rate, etc. The higher the exercise score, accuracy rate or the lower the error rate, the better the user's mastery of the knowledge points of the first knowledge question set extracted in this round.

[0050] Specifically, after receiving the initial response answer set, the server obtains the practice result data by calculating the semantic similarity between the initial response answer set and the reference answers corresponding to the knowledge questions in the first knowledge question set, so as to subsequently combine the client's practice result data to dynamically extract knowledge questions, so that the extracted knowledge questions have a strong correlation with the client's practice effects, which is conducive to improving the knowledge practice effect.

[0051] Step 204, based on the exercise result data, re-select questions from the knowledge question bank to obtain a second knowledge question set, and perform the next round of knowledge exercises on the second knowledge question set until the exercise end condition is met.

[0052] The second set of knowledge questions is a set of knowledge questions re-extracted from the knowledge question bank, and the basis for the re-extraction is the exercise result data of the previous round of knowledge exercises. The exercise end condition is information used to determine the end of the knowledge exercise process. For example, the exercise end condition can be the number of knowledge exercise rounds, the total number of knowledge questions in the knowledge exercise, the knowledge exercise end instruction, etc.

[0053] Specifically, the server re-extracts questions from the knowledge question bank based on the practice result data to obtain a second set of knowledge questions, thereby realizing dynamic and cyclic extraction of knowledge questions in combination with the client's knowledge practice situation, ensuring that the second set of knowledge questions can accurately reflect the client's knowledge practice situation, and performing multiple rounds of knowledge practice on the second knowledge question set, thereby making the client's knowledge practice more targeted and greatly improving the knowledge practice effect.

[0054] In a specific embodiment, Figure 4 As shown, based on the exercise result data, re-select questions from the knowledge question bank to obtain a second knowledge question set, and perform the next round of knowledge exercises on the second knowledge question set to complete a knowledge training. The process may include the following steps:

[0055] Step 401, re-extracting from the knowledge question bank according to the exercise result data to obtain a second knowledge question set;

[0056] Step 402, sending the second knowledge question set to the client, and determining updated practice result data according to the second response answer set for the second knowledge question set received from the client;

[0057] Step 403, re-extract from the knowledge question bank based on the updated exercise result data to obtain an updated second knowledge question set, and return to the step of sending the updated second knowledge question set to the client to continue executing until the exercise end condition is met, thereby completing this knowledge exercise.

[0058] Specifically, the probability of drawing knowledge questions for each type of knowledge point can be determined based on the practice result data, and questions are re-drawn from the knowledge question bank according to the probability of drawing knowledge questions to obtain a second set of knowledge questions, and then the second set of knowledge questions is sent to the client, and the updated practice result data is determined based on the second response answer set received from the client, the second set of knowledge questions is updated based on the updated practice result data, and step 402 is executed repeatedly until the practice end condition is met, thereby realizing multiple rounds of knowledge practice in one knowledge practice process, and because the questions drawn in multiple rounds of knowledge practice are dynamically and cyclically drawn based on the practice result data of the previous round, the knowledge questions drawn for knowledge practice have a strong correlation with the client's mastery of the knowledge, which greatly improves the effect of knowledge practice compared to the existing method of drawing all knowledge questions for practice at one time.

[0059] The knowledge practice method provided in the embodiment of the present application first obtains a knowledge practice topic from a knowledge practice instruction, and extracts multiple knowledge questions from a knowledge question bank that matches the knowledge practice topic, which can ensure that the knowledge questions for the knowledge practice are related to the knowledge practice topic, so that the knowledge questions for the knowledge practice are targeted; then, an initial response answer set of a first knowledge question set is received, and based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, practice result data for a round of knowledge practice is determined, and based on the practice result data, questions are re-drawn from the knowledge question bank to obtain a second knowledge question set, thereby realizing dynamic and cyclic extraction of knowledge questions combined with practice conditions, realizing real-time dynamic question extraction, ensuring that the re-extracted knowledge question set can accurately reflect the knowledge practice situation, and performing multiple rounds of practice on the dynamically extracted knowledge question set, thereby making the knowledge practice more targeted and greatly improving the knowledge practice effect.

[0060] In one embodiment, each knowledge question in the knowledge question bank corresponds to a category knowledge point, and re-extracting questions from the knowledge question bank based on the practice result data to obtain a second knowledge question set includes: determining a total category knowledge point set corresponding to all knowledge questions in the knowledge question bank, and a first category knowledge point set corresponding to all knowledge questions in the first knowledge question set; determining a target category knowledge point set based on the total category knowledge point set, the first category knowledge point set and the practice result data; and re-extracting at least one knowledge question from the knowledge question bank based on the practice result data and the target category knowledge point set to obtain a second knowledge question set.

[0061] Among them, each knowledge question in the knowledge question bank corresponds to a category knowledge point. Exemplarily, in the knowledge question bank matching the product theme, there are knowledge questions of 40 category knowledge points, and each category knowledge point corresponds to at least one knowledge question.

[0062] The total category knowledge point set is the set of category knowledge points corresponding to all knowledge questions in the knowledge question bank. The combination of category knowledge corresponding to each knowledge question in the knowledge question bank can be used as the total category knowledge point set. The first category knowledge point set is the set of category knowledge points corresponding to all knowledge questions in the first knowledge question set, that is, the set of category knowledge points corresponding to the knowledge questions extracted in the previous round of knowledge exercises.

[0063] The target category knowledge point set is determined based on the total category knowledge point set, the first category knowledge point set and the exercise result data, and is used as the basis for re-drawing questions from the knowledge question bank in the next round of knowledge exercise. For example, the target category knowledge point set includes 5 category knowledge points, namely A, B, C, D, and E. In the next round of knowledge exercise, knowledge questions about the 5 category knowledge points in the target category knowledge point set can be extracted, or questions about some category knowledge points can be extracted from the 5 category knowledge points, such as knowledge questions about 4 category knowledge points, namely A, B, C, and D.

[0064] Specifically, at least one knowledge question is re-extracted from the knowledge question bank based on the practice result data and the target category knowledge point set to obtain a second knowledge question set. It can be understood that in this embodiment, since the target category knowledge point set is determined based on the total category knowledge point set, the first category knowledge point set and the practice result data, on the one hand, based on the total category knowledge point set and the first category knowledge point set, comprehensive coverage of subsequent category knowledge points is taken into account. On the other hand, based on the first category knowledge point set and the practice result data, the mastery of the first category knowledge points in the previous round of knowledge practice is taken into consideration, thereby ensuring that the determined target category knowledge point set can not only meet the comprehensive coverage conditions of category guidance, but also ensure that the category knowledge points are highly targeted to users, thereby improving the effectiveness of the target category knowledge point set. Furthermore, based on the practice result data and the target category knowledge point set, at least one knowledge question is re-extracted from the knowledge question bank to obtain a second knowledge question set. Since the target category knowledge point set has high effectiveness and rationality, the practice result data and the target category knowledge point set are continued to be combined to re-extract questions from the knowledge question bank, thereby improving the effectiveness and rationality of the second knowledge question set, which is conducive to improving the effectiveness of subsequent knowledge practice.

[0065] In one embodiment, determining the target category knowledge point set based on the total category knowledge point set, the first category knowledge point set and the exercise result data includes: when the total category knowledge point set is the same as the first category knowledge point set, determining the total category knowledge point set as the target category knowledge point set; when the total category knowledge point set is different from the first category knowledge point set, determining the target category knowledge point set from the total category knowledge point set based on the exercise result data.

[0066] Specifically, in some cases, when the total category knowledge point set is the same as the first category knowledge point set, it indicates that the category knowledge points corresponding to the knowledge questions in the first knowledge question set extracted in the previous round of knowledge exercises cover all category knowledge points in the knowledge question bank. Therefore, there is no need to screen the category knowledge points. By determining the target category knowledge point set as the total category knowledge point set, the target category knowledge point set can cover all category knowledge points and ensure the comprehensiveness of subsequent knowledge exercises.

[0067] In other cases, when the total category knowledge point set is different from the first category knowledge point set, it indicates that the category knowledge points corresponding to the knowledge questions in the first knowledge question set extracted in the previous round of knowledge exercises do not cover all category knowledge points in the knowledge question bank, that is, there is at least one category knowledge point whose knowledge questions have not been extracted. Therefore, combined with the exercise result data, the target category knowledge point set can be determined from the total category knowledge point set in a targeted manner, thereby improving the comprehensiveness and effectiveness of knowledge exercises.

[0068] In one embodiment, determining a target category knowledge point set from the total category knowledge point set based on the exercise result data includes: determining the difference between the total category knowledge point set and the first category knowledge point set to obtain a second category knowledge point set; determining a target first category knowledge point set from the first category knowledge point set based on the exercise result data; and determining the target first category knowledge point set and the second category knowledge point set as the target category knowledge point set.

[0069] The second category knowledge point set is the difference between the total category knowledge point set and the first category knowledge point set, that is, the set of category knowledge points that were not extracted in the previous round of knowledge practice.

[0070] The target first category knowledge point set is a set of first category knowledge points selected from the first category knowledge point set according to the practice result data. Exemplarily, the first category knowledge point set includes 4 category knowledge points, namely, A, B, C, and D, and the corresponding practice result data are respectively the accuracy meets the condition, the accuracy meets the condition, the accuracy does not meet the condition, and the accuracy does not meet the condition, indicating that the user is proficient in the A and B category knowledge points and is not proficient in the C and D category knowledge points. In order to help the user find problems, the C and D category knowledge points can be selected from the 4 category knowledge points of A, B, C, and D as the target first category knowledge point set, so that the user can continue to practice the category knowledge points in the target first category knowledge point set in the next round of knowledge practice, helping the user to improve the mastery of the C and D category knowledge points.

[0071] Specifically, the target first category knowledge point set and the second category knowledge point set are determined as the target category knowledge point set. Since the second category knowledge point set is a set of category knowledge points that have not been drawn, using the second category knowledge point set as the target category knowledge point set can ensure the comprehensiveness of the knowledge practice. Since the target first category knowledge point set is determined based on the practice result data, it is possible to ensure that the target first category knowledge points are targeted. In this embodiment, the target first category knowledge point set and the second category knowledge point set are determined as the target category knowledge point set, so that the target category knowledge point set can meet the requirements of comprehensive coverage of knowledge practice, while also meeting the requirements of targeted knowledge practice, which is conducive to improving the effect of knowledge practice.

[0072] In one embodiment, based on the exercise result data and the target category knowledge point set, at least one knowledge question is re-extracted from the knowledge question bank to obtain a second knowledge question set, including: determining the extraction weights of knowledge questions of each category knowledge point in the target category knowledge point set according to the exercise result data and the target category knowledge point set; and re-extracting at least one knowledge question from the knowledge question bank according to the extraction weights to obtain the second knowledge question set.

[0073] Among them, the extraction weight refers to the weighted value of the probability of a knowledge question of a category knowledge point being drawn. Generally, the probability of a knowledge question of a category knowledge point being drawn is the same. Therefore, the higher the extraction weight, the higher the probability of a knowledge question of the corresponding category knowledge point being drawn.

[0074] Specifically, based on the exercise result data and the target category knowledge point set, the extraction weights of the knowledge questions of each category knowledge point in the target category knowledge point set are determined, and the implementation process may be: first, based on the exercise result data, the first result data of each first category knowledge point in the first category knowledge point set are counted, wherein the first result data is comprehensive result data for characterizing a category knowledge point, such as a comprehensive exercise score, a comprehensive accuracy rate, a comprehensive error rate, etc. Exemplarily, the exercise result data is an exercise score, and the first result data may be the mean, median, minimum value, maximum value, etc. of all exercise scores of a category knowledge point. Then, based on the first result data, category knowledge points corresponding to the first result data that meet the preset conditions are selected as the mastered category knowledge points, and the remaining category knowledge points are selected as unmastered category knowledge points; then, if there are unmastered category knowledge points and mastered category knowledge points in the target category knowledge points, the knowledge question extraction weight of the unmastered category knowledge points is set to be higher than the knowledge question extraction weight of the mastered category knowledge points. At the same time, for the extraction weight of the knowledge questions of the remaining category knowledge points in the target category knowledge points, a preset number of category knowledge points can be randomly selected and a higher weight can be set, or the same extraction weight can be set for all of them, and there is no restriction here.

[0075] It can be understood that in this embodiment, the exercise result data and the target category knowledge point set are combined to determine the extraction weights of the knowledge questions of each category knowledge point, so that the extracted second knowledge question set is closely related to the practice situation of the previous round of knowledge exercises, making the second knowledge question set more targeted to users. It not only realizes the evaluation and analysis of the mastery of each category knowledge point, but also makes the second knowledge question set of the next round combined with the dynamic changes of the historical knowledge practice situation, which can provide users with more comprehensive and targeted knowledge exercises, and is conducive to further improving the effect of knowledge exercises.

[0076] In a specific embodiment, based on the exercise result data and the target category knowledge point set, re-extracting at least one knowledge question from the knowledge question bank to obtain a second knowledge question set may include: determining a first extraction weight of the knowledge questions of each target first category knowledge point in the target first category knowledge point set based on the exercise result data; determining a second extraction weight of the knowledge questions of each second category knowledge point in the second category knowledge point set; and re-extracting at least one knowledge question from the knowledge question bank according to each target first category knowledge point and each second category knowledge point, and each corresponding first extraction weight and second extraction weight, to obtain a second knowledge question set.

[0077] As a possible implementation method, according to the practice result data, the first result data of each first category knowledge point in the first category knowledge point set are counted, and according to the first result data, the category knowledge points corresponding to the first result data that meet the preset conditions are selected as the mastered category knowledge points, and the remaining category knowledge points are selected as the unmastered category knowledge points. The first extraction weight of the knowledge questions corresponding to the mastered category knowledge points in the target first category knowledge points is set to be lower than the preset value (for example, the preset value is 50%), and the first extraction weight of the knowledge questions corresponding to the mastered category knowledge points in the target first category knowledge points is set to be higher than the preset value, so as to ensure that the knowledge questions corresponding to the unmastered category knowledge points can be repeatedly extracted, which helps users improve their mastery of the category knowledge points, and ensures that the knowledge questions corresponding to the mastered category knowledge points will not be repeatedly extracted, thereby improving the efficiency of knowledge practice.

[0078] As a possible implementation method, the second extraction weight of the knowledge questions of each second category knowledge point in the second category knowledge point set is determined. The specific implementation method may be: the second extraction weight of the knowledge questions of each second category knowledge point is set to be higher than a preset value (for example, the preset value is 50%), so as to ensure that the knowledge questions corresponding to the category knowledge points that have not been extracted can be extracted, thereby improving the coverage of the category knowledge points of the knowledge exercise and making the knowledge exercise more comprehensive.

[0079] In one embodiment, the knowledge practice instruction includes a target practice mode; the knowledge questions and reference answers in the knowledge question bank are a default practice mode; before receiving the initial response answer set of the first knowledge question set and determining the practice result data of a round of knowledge practice based on the initial response answer set and the reference answers corresponding to each knowledge question in the first knowledge question set, it also includes: when the target practice mode is different from the default practice mode, converting the first knowledge question set of the default practice mode into a first knowledge question set consistent with the target practice mode.

[0080] The target practice mode and the default practice mode may be one of a video practice mode, a voice practice mode or a text practice mode.

[0081] Specifically, the knowledge practice method of this embodiment supports multiple practice modes, video mode, voice mode or text mode. Correspondingly, the knowledge practice instructions include a target practice mode. When the target practice mode is different from the default practice mode, in order to improve the knowledge practice efficiency of the client, before receiving the initial response answer set of the first knowledge question set, that is, before sending the first knowledge question set to the user, the first knowledge question set of the default practice mode is converted into a first knowledge question set that matches the target practice mode, so that the user can continue to use the target practice mode for knowledge practice.

[0082] In one embodiment, the knowledge practice instruction includes a target practice mode; the initial response answer set of the first knowledge question set is received, and based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, the practice result data of a round of knowledge practice is determined.

[0083] In order to improve the efficiency of determining the practice result data, when it is detected that the response answers in the initial response answer set are the target practice mode, and the target practice mode is different from the default practice mode corresponding to the knowledge question bank, the response answers in the initial response answer set are converted into an initial response answer set matching the default practice mode, thereby ensuring that the server can determine the practice result data based on the initial response answer set matching the default practice mode and the reference answers corresponding to the knowledge questions in the first knowledge question set, thereby realizing knowledge practice in multiple practice modes and improving the efficiency of knowledge practice.

[0084] In a specific implementation, taking the target practice mode as the voice practice mode and the default practice mode as the text practice mode as an example, the first knowledge question set is converted into the first knowledge question set in voice form, and the first knowledge question set in voice form is sent to the customer, and the initial response answer set is converted into the initial response answer set in text form, and the practice result data is determined according to the initial response answer set in text form and the reference answers corresponding to the knowledge questions in the first knowledge question set. Figure 8 As shown, it is a schematic diagram of the knowledge practice process in the voice practice mode, that is, through the text-to-speech conversion (TextToSpeech, TTS) service, the first knowledge question set is converted into the first knowledge question set in voice form, and through the automatic speech recognition (Automatic Speech Recontion, ASR) service, the initial response answer set is converted into the initial response answer set in text form. In this embodiment, by providing multiple knowledge practice modes, such as voice practice mode and text mode, it is beneficial to improve the convenience of client practice and further improve the practice effect.

[0085] In one embodiment, before extracting knowledge questions from the knowledge question bank matching the knowledge practice theme to obtain a first set of knowledge questions, the method also includes: acquiring multimodal knowledge data and converting the multimodal knowledge data into textual knowledge data in text form; based on the textual knowledge data, using a pre-trained large language model for analysis to generate a knowledge question bank; and / or, based on a preset interface, acquiring target knowledge data corresponding to the knowledge practice theme; constructing a knowledge question bank based on external knowledge data to obtain a knowledge question bank.

[0086] Among them, multimodal knowledge data refers to knowledge data in different forms of expression, such as video knowledge data, audio knowledge data, text knowledge data, etc. A pre-trained large language model (LLM) is an artificial intelligence model designed to understand and generate human language, such as GPT-4, PaLM, Galactica, or ChatGLM. The preset interface is an interface for obtaining external data sources.

[0087] Specifically, one way to generate a knowledge question bank can be to obtain multimodal knowledge data through a web crawler or from a historical database of a knowledge practice system, and convert the multimodal knowledge data into textual knowledge data in text form, and then use a pre-trained large language model to analyze the textual knowledge data to generate a knowledge question bank. It can be understood that in this embodiment, the richness and integrity of the knowledge data are ensured by using multimodal knowledge data, and at the same time, the pre-trained large language model is used for analysis, so that the target knowledge base can be constructed without the help of a large number of training samples, thereby improving the construction efficiency of the knowledge base.

[0088] Another way to generate the knowledge question bank may be: according to a preset interface, the target knowledge data associated with the exercise theme is synchronized regularly through a scheduled task, and then the target knowledge data is analyzed and sorted to construct the knowledge question bank. It can be understood that in this embodiment, by acquiring the target knowledge data and constructing the knowledge question bank, the knowledge question bank can be updated in real time, which improves the coverage of knowledge points in the knowledge question bank and further improves the accuracy and completeness of the knowledge question bank.

[0089] Another way to generate a knowledge question bank can be: a combination of the above two generation methods, that is, a knowledge question bank obtained by analyzing a pre-trained large language model and a knowledge question bank constructed by target knowledge data, which further improves the accuracy and completeness of the knowledge question bank.

[0090] It is worth noting that an editing function for the knowledge question bank can also be provided, so as to edit and correct the knowledge question bank through an interface and / or an interface, thereby further improving the accuracy of the knowledge question bank.

[0091] In one embodiment, based on the textual knowledge data, a pre-trained large language model is used for analysis to generate a knowledge question bank, including: constructing a target prompt word based on the textual knowledge data; inputting the target prompt word into the pre-trained large language model for analysis to obtain a knowledge question bank.

[0092] The target prompt word refers to the prompt obtained after reconstructing the text knowledge data, which is used as the input of the large language model.

[0093] Specifically, the text knowledge data can be processed in pieces, and the length of the token (number of characters) of the text knowledge data can be controlled by the piece processing, and the text knowledge data processed in pieces can be filled into a pre-configured prompt template, and the context obtained by the construction is the target prompt word, wherein the prompt template can be configured according to the classification dimension of the text knowledge data. Then the target prompt word is input into a pre-trained large language model for analysis and reasoning to obtain a knowledge question bank. It can be understood that in this embodiment, the text knowledge data is processed in pieces to improve the reasoning speed of the large language model, and then the text knowledge data processed in pieces is filled into a pre-configured prompt template. Since the prompt template is configured according to the classification dimension of the text knowledge data, the context obtained by the construction is knowledge that can be understood by the large language model, which improves the accuracy of the target prompt word, thereby improving the accuracy and generation efficiency of the knowledge question bank.

[0094] In one embodiment, a target prompt word is input into a pre-trained large language model for analysis and reasoning to obtain a knowledge question bank, including: inputting the target prompt word into the pre-trained large language model to obtain an initial question-answer pair output by the pre-trained large language model, wherein the initial question-answer pair includes an initial question and an initial answer; scoring the initial question-answer pair to obtain a first score for the initial question-answer pair; determining a target question-answer pair corresponding to the initial question pair based on the first score and the initial question-answer pair, taking the initial question in the target question-answer pair as the knowledge question, and taking the initial answer in the target question-answer pair as a reference answer corresponding to the knowledge question; and generating the knowledge question bank based on all the acquired knowledge questions and the reference answers corresponding to the knowledge questions.

[0095] Among them, the first score is used to characterize the indicator data of the matching degree between the initial question and the initial answer in the initial question and answer pair. The standard answer corresponding to the question can be obtained from the preset standard knowledge base, and the semantic similarity between the standard answer and the initial answer is calculated as the first score. The matching degree score between the knowledge question and the initial answer in the initial question and answer pair can also be determined by a pre-trained question and answer pair scoring model. The question and answer pair scoring model can be a bag of words (BOW) model, a large language model, etc.

[0096] Specifically, scoring the initial question-answer pair output by the pre-trained large language model can be performed by inputting the initial question-answer pair into a preset question-answer pair scoring model, obtaining a first score output by the preset question-answer pair scoring model, determining the target question-answer pair corresponding to the initial question pair based on the first score and the initial question-answer pair, using the initial question in the target question-answer pair as the knowledge question, using the initial answer in the target question-answer pair as the reference answer corresponding to the knowledge question, and the combination of all the knowledge questions and the reference answers corresponding to the knowledge questions obtained is the knowledge question bank. It can be understood that in this embodiment, by scoring the initial question-answer pair output by the pre-trained large language model, the accuracy of the initial question-answer pair is evaluated, and based on the evaluation results, the initial question-answer pair is analyzed and reasoned, which improves the accuracy of the knowledge question bank compared to directly using the initial question-answer pair as the knowledge question bank.

[0097] In one embodiment, the target question-answer pair corresponding to the initial question pair is determined based on the first score and the initial question-answer pair, including: if the first score meets the scoring condition, the initial question-answer pair is determined as the target question-answer pair, the initial question in the target question-answer pair is used as the knowledge question, and the initial answer in the target question-answer pair is used as the reference answer corresponding to the knowledge question; if the first score does not meet the scoring condition, knowledge data is retrieved based on the initial question in the initial question-answer pair to obtain updated knowledge data; updated target prompt words are reconstructed based on the updated knowledge data, and the updated target prompt words are input into the pre-trained large language model to obtain an updated question-answer pair; the updated question-answer pair is scored to obtain a second score for the updated question-answer pair; if the second score meets the scoring condition, the updated question-answer pair is determined as the target question-answer pair.

[0098] Among them, the scoring condition is used to measure the critical value of the matching degree between the question and the answer in the question-answer pair, such as 80%, 90%, etc.

[0099] The second score is used to characterize the indicator data of the matching degree between the updated question and the updated answer in the updated question-answer pair, and the second score can be determined in the same way as the first score.

[0100] Specifically, when the first score meets the scoring conditions, it indicates that the matching degree between the initial question and the initial answer in the initial question-answer pair meets the conditions, that is, the initial answer is the correct answer to the initial question. Then the initial question-answer pair is determined as the target question-answer pair, the initial question in the target question-answer pair is used as the knowledge question, the initial answer in the target question-answer pair is used as the reference answer corresponding to the knowledge question, and all knowledge questions and the reference answers corresponding to the knowledge questions are used to generate a knowledge question bank. Figure 5As shown in the figure, it is a schematic diagram of the target prompt word reasoning process. After obtaining the initial question-answer pair, the preset question-answer pair is further scored using the preset question-answer pair evaluation model. When the score of the preset question-answer pair meets the scoring condition, the initial question-answer pair is determined to be the target question-answer pair. The initial question-answer pair output by the pre-trained large language model is evaluated and analyzed by the preset question-answer pair scoring model, and the initial question-answer pair whose score meets the scoring condition is used as the knowledge question bank, which realizes further verification of the initial question-answer pair and further improves the accuracy of the knowledge question bank.

[0101] When the first score does not meet the scoring conditions, it indicates that the initial question and the initial answer in the initial question and answer pair do not meet the conditions, that is, the initial answer is not an accurate answer to the initial question, then the knowledge data is retrieved according to the initial question in the initial question and answer pair to obtain updated knowledge data, and the context is reconstructed based on the updated knowledge data. Among them, the context is reconstructed according to the updated knowledge data. The above-mentioned method of constructing target prompt words can be used to obtain updated target prompt words, and the updated target prompt words are input into the pre-trained large language model to obtain an updated question and answer pair, thereby realizing the update processing of the inaccurate initial question and answer pair.

[0102] After obtaining the updated question-answer pair, continue to use the preset question-answer pair scoring model to score the updated question-answer pair to obtain the corresponding second score. When the second score meets the scoring conditions, it indicates that the matching degree between the updated question and the updated answer in the updated question-answer pair meets the conditions, that is, the updated answer is the accurate answer to the updated question. Then the updated question-answer pair is determined as the target question-answer pair to obtain the knowledge question base. When the second score does not meet the scoring conditions, the updated question-answer pair is deleted.

[0103] like Figure 6 As shown, it is another schematic diagram of the target prompt word reasoning process. After obtaining the initial question and answer pair, the preset question and answer pair evaluation model is used to continue to score the preset question and answer pair. When the score of the preset question and answer pair does not meet the scoring conditions, the knowledge data is retrieved according to the initial question in the initial question and answer pair. After obtaining the updated knowledge data, the context of the updated knowledge data is reconstructed to obtain the updated target prompt word, and the updated target prompt word is continued to be reasoned through the pre-trained large language model and the question and answer pair evaluation model, thereby realizing the secondary evaluation and analysis of the initial question and answer pair. It can be understood that when analyzing and reasoning the target prompt word, this embodiment evaluates and analyzes the initial question and answer pair output by the pre-trained large language model, and retrieves knowledge data for the initial question and answer that does not meet the scoring conditions, so as to further enhance the context information, thereby obtaining a more accurate answer, and further improving the accuracy of the knowledge question bank.

[0104] In one embodiment, the method further includes: obtaining a sample knowledge data set, and constructing a sample prompt word set based on the sample knowledge data set; inputting the sample prompt word set into a preset instruction language model to obtain an instruction fine-tuning data set; using a preset model fine-tuning method, and using the instruction fine-tuning data set to fine-tune the preset large language model to obtain a pre-trained large language model.

[0105] The sample knowledge data set can be obtained from the knowledge material library. The preset instruction language model is a large language model used for fine-tuning instructions, for example, a base large language model (Base Large Language Model, Base LLM). The preset model fine-tuning method is a pre-set model fine-tuning method, such as the Freeze method, the PT method (including P Tuning and P Tuning V2) or the non-full-quantity fine-tuning method (Low-Rank Adaptation, Lora).

[0106] Specifically, a sample prompt word set is input into a preset instruction language model to obtain an instruction fine-tuning data set, and a preset model fine-tuning method is adopted to fine-tune the preset large language model using the instruction fine-tuning data set to obtain a pre-trained large language model. In this embodiment, the preset large language model is fine-tuned and trained using the instruction fine-tuning data set, that is, the instruction fine-tuning of the preset large language model is achieved by specially training the preset large language model, so as to better understand the sample knowledge data set and give answers that meet the instructions, thereby improving the precision and accuracy of the pre-trained large language model.

[0107] In one embodiment, multimodal knowledge data includes document knowledge data, audio knowledge data and video knowledge data; acquiring multimodal knowledge data and converting the multimodal knowledge data into text knowledge data in text form, including: using a preset optical character recognition model to recognize document knowledge data to obtain first knowledge data; extracting audio data from video knowledge data; recognizing audio knowledge data and audio data to obtain second knowledge data; and determining text knowledge data based on the first knowledge data and the second knowledge data.

[0108] Specifically, for document knowledge data, a preset optical character recognition (OCR) model is used for recognition to obtain the corresponding first knowledge data; for video knowledge data, the audio data in the video knowledge data is extracted, and a preset automatic speech recognition (ASR) model can be used to recognize the audio knowledge data and the audio data to obtain the second knowledge data; the combination of the first knowledge data and the second knowledge data is the text knowledge data. Figure 7In this embodiment, by using corresponding data recognition models to analyze and recognize multimodal knowledge data, the multimodal knowledge data is converted into textual knowledge data, so as to improve the efficiency of knowledge question bank construction based on textual knowledge data.

[0109] In one embodiment, the preset optical character recognition model includes a first character recognition model, a second character recognition model and a third character recognition model connected in parallel, the first character recognition model, the second character recognition model and the third character recognition model are models obtained by training the character recognition model with samples in text form, samples in table form and samples in image format, respectively; the preset optical character recognition model is used to recognize document knowledge data to obtain first knowledge data, including: converting the document knowledge data into knowledge data in a preset format; performing version analysis on the knowledge data in the preset format to identify text data in text format, table data in table format and image data in image format; using the first character recognition model to recognize text data to obtain first recognition data; using the second character recognition model to recognize table data to obtain second recognition data; using the third character recognition model to recognize image data to obtain third recognition data; determining the first knowledge data based on the first recognition data, and / or the second recognition data, and / or the third recognition data.

[0110] The preset format may be a PDF format. The preset optical character recognition model includes a first character recognition model, a second character recognition model, and a third character recognition model connected in parallel, wherein the first character recognition model, the second character recognition model, and the third character recognition model are respectively obtained by training the character recognition model with samples in text form, samples in table form, and samples in image format.

[0111] Specifically, document knowledge data is converted into knowledge data in PDF format; version analysis is performed on the knowledge data in PDF format to identify text data in text format, table data in table format, and image data in image format; a first character recognition model is used to recognize text data to obtain first recognition data; a second character recognition model is used to recognize table data to obtain second recognition data; a third character recognition model is used to recognize image data to obtain third recognition data, and at least one of the first recognition data, the second recognition data, and the third recognition data is used as the first knowledge data. It can be understood that in this embodiment, for different forms of document knowledge data, corresponding OCR models trained for the respective different forms of document knowledge data are used for recognition, which improves the recognition efficiency of the character recognition model compared to the use of a general OCR model.

[0112] In one embodiment, the second character recognition model includes multiple second character recognition sub-models; the second character recognition model is used to recognize the table data to obtain second recognition data, including: extracting feature vector data corresponding to the table data; using the feature vector data as input of each second character recognition sub-model to obtain multiple recognition results output by each second character recognition sub-model; using weighted voting to select a target recognition result from the multiple recognition results, and using the target recognition result as the second recognition data.

[0113] Specifically, word embedding can be used to vectorize the table data to extract feature vector data corresponding to the table data, and the feature vector data is used as the input of each second character recognition sub-model to obtain multiple recognition results output by each second character recognition sub-model. A weighted voting method is used to select a target recognition result from the multiple recognition results, and the target recognition result is used as the second recognition data. In this embodiment, the accuracy of the second recognition data is improved by voting by multiple models.

[0114] In one embodiment, after re-drawing questions from the knowledge question bank based on the practice result data to obtain a second knowledge question set, and performing the next round of knowledge practice on the second knowledge question set until the practice end condition is met, it also includes: determining the category practice result data of each category knowledge point in each round of knowledge practice based on the practice result data of each round of knowledge practice; counting the comprehensive practice result data of each category knowledge point according to the category practice result data of each category knowledge point in each round of knowledge practice; and generating a radar chart for each category knowledge point according to the comprehensive practice result data of each category knowledge point.

[0115] The comprehensive practice result data is statistical practice data for representing a category of knowledge points in a round of knowledge practice, such as statistical scores, statistical accuracy, statistical error rate, etc. The comprehensive practice result data for each category of knowledge points is statistically analyzed, and the statistical method includes but is not limited to statistically analyzing the mean, median, minimum, maximum, etc. of all practice scores of a category of knowledge points in a round of knowledge practice. Exemplarily, the statistical practice data is statistical scores, and the comprehensive practice result data can be the mean, median, minimum, maximum, etc. of the category practice result data in a round of knowledge practice.

[0116] Specifically, after the end of this exercise, based on the exercise result data of each round of knowledge exercise, the comprehensive exercise score of each category of knowledge point in each round of knowledge exercise is determined, and according to the category exercise result data of each category of knowledge point in each round of knowledge exercise, the comprehensive exercise result data of each category of knowledge point is counted; based on the comprehensive exercise result data of each category of knowledge point, a radar chart of each category of knowledge point is generated, and the comprehensive exercise result data of each category of knowledge point in the knowledge exercise process is presented through the radar chart, so that the client can have a more intuitive understanding of the degree of mastery of each category of knowledge point, further improving the effect of knowledge exercise.

[0117] It should be understood that, although each step in the flowcharts involved in the above embodiments is displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiment of the present application also provides a knowledge practice device for implementing the above-mentioned knowledge practice method, and a knowledge practice device for implementing the above-mentioned knowledge practice method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more knowledge practice devices provided below can refer to the limitations of the knowledge practice method above, and the specific limitations will not be repeated here.

[0119] This embodiment also provides a knowledge training device, which can be integrated into a terminal device. Fig. 9 As shown, the knowledge practice device 900 may include:

[0120] An acquisition module 901 is used to respond to a knowledge practice instruction and acquire a knowledge practice topic from the knowledge practice instruction;

[0121] The question extraction module 902 is used to extract knowledge questions from the knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer;

[0122] A determination module 903 is used to receive an initial response answer set of the first knowledge question set, and determine the exercise result data of a round of knowledge exercise based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set;

[0123] The practice module 904 is used to re-select questions from the knowledge question bank based on the practice result data to obtain a second knowledge question set, and perform the next round of knowledge practice on the second knowledge question set until the practice end condition is met.

[0124] The above-mentioned knowledge practice device first obtains the knowledge practice topic from the knowledge practice instruction, and extracts multiple knowledge questions from the knowledge question bank matching the knowledge practice topic, which can ensure that the knowledge questions for the knowledge practice are related to the knowledge practice topic, so that the knowledge questions for the knowledge practice are targeted; then, the initial response answer set of the first knowledge question set is received, and based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, the practice result data of a round of knowledge practice is determined, and based on the practice result data, questions are re-drawn from the knowledge question bank to obtain the second knowledge question set, thereby realizing dynamic and cyclic extraction of knowledge questions combined with the practice situation, realizing real-time dynamic question extraction, ensuring that the re-extracted knowledge question set can accurately reflect the knowledge practice situation, and performing multiple rounds of practice on the dynamically extracted knowledge question set, thereby making the knowledge practice more targeted and greatly improving the knowledge practice effect.

[0125] In some embodiments, each knowledge question in the knowledge question bank corresponds to a category knowledge point, and the practice module 904 is specifically used to:

[0126] Determine a total category knowledge point set corresponding to all knowledge questions in the knowledge question bank, and a first category knowledge point set corresponding to all knowledge questions in the first knowledge question set;

[0127] Determine a target category knowledge point set according to the total category knowledge point set, the first category knowledge point set and the exercise result data;

[0128] Based on the practice result data and the target category knowledge point set, at least one knowledge question is re-extracted from the knowledge question bank to obtain a second knowledge question set.

[0129] In some embodiments, the exercise module 904 is further configured to:

[0130] When the total category knowledge point set is the same as the first category knowledge point set, determining the total category knowledge point set as the target category knowledge point set;

[0131] When the total category knowledge point set is different from the first category knowledge point set, a target category knowledge point set is determined from the total category knowledge point set according to the exercise result data.

[0132] In some embodiments, the exercise module 904 is further configured to:

[0133] The step of determining a target category knowledge point set from the total category knowledge point set according to the exercise result data comprises:

[0134] Determine a difference between the total category knowledge point set and the first category knowledge point set to obtain a second category knowledge point set;

[0135] Determining a target first category knowledge point set from the first category knowledge point set according to the exercise result data;

[0136] The target first category knowledge point set and the second category knowledge point set are determined as the target category knowledge point set.

[0137] In some embodiments, the exercise module 904 is further configured to:

[0138] Determining the extraction weight of knowledge questions of each category knowledge point in the target category knowledge point set according to the exercise result data and the target category knowledge point set;

[0139] According to the extraction weight, at least one knowledge question is re-extracted from the knowledge question bank to obtain the second knowledge question set.

[0140] In some embodiments, the knowledge practice instruction includes a target practice mode; the knowledge questions and reference answers in the knowledge question bank are default practice modes; the knowledge practice device also includes:

[0141] A first conversion module is configured to convert a first knowledge question set of the default practice mode into a first knowledge question set consistent with the target practice mode when the target practice mode is different from the default practice mode.

[0142] In some embodiments, the knowledge practice instruction includes a target practice mode; the determination module 903 is specifically used to:

[0143] When the response answers in the initial response answer set are the target practice mode, and the target practice mode is different from the default practice mode corresponding to the knowledge question bank, converting the response answers in the initial response answer set into an initial response answer set matching the default practice mode;

[0144] According to the initial response answer set consistent with the default practice mode and the reference answers corresponding to the knowledge questions in the first knowledge question set, the practice result data of a round of knowledge practice is determined. In some embodiments, the knowledge practice device further includes:

[0145] A second conversion module is used to obtain multimodal knowledge data and convert the multimodal knowledge data into textual knowledge data in text form;

[0146] A generation module is used to generate a knowledge question bank based on text knowledge data by using a pre-trained large language model for analysis; and / or,

[0147] A first acquisition module, used to acquire target knowledge data corresponding to the knowledge practice topic based on a preset interface;

[0148] A construction module is used to construct a knowledge question base based on the target knowledge data to obtain the knowledge question base.

[0149] In some embodiments, the generating module is specifically used to:

[0150] Construct target prompt words based on text knowledge data;

[0151] The target prompt words are input into the pre-trained large language model for analysis and reasoning to obtain a knowledge question base.

[0152] In some embodiments, the generating module is further configured to:

[0153] Input the target prompt word into the pre-trained large language model to obtain the initial question-answer pair output by the pre-trained large language model, where the initial question-answer pair includes an initial question and an initial answer;

[0154] Scoring the initial question-answer pair to obtain a first score for the initial question-answer pair;

[0155] Determine, according to the first score and the initial question-answer pair, a target question-answer pair corresponding to the initial question pair, use the initial question in the target question-answer pair as the knowledge question, and use the initial answer in the target question-answer pair as a reference answer corresponding to the knowledge question;

[0156] The knowledge question bank is generated based on all the acquired knowledge questions and the reference answers corresponding to the knowledge questions.

[0157] In some embodiments, the generating module is further configured to:

[0158] If the first score satisfies the score condition, determining the initial question-answer pair as the target question-answer pair;

[0159] If the first score does not satisfy the score condition, retrieving knowledge data according to the initial question in the initial question-answer pair to obtain updated knowledge data;

[0160] constructing an updated target prompt word based on the updated knowledge data, and inputting the updated target prompt word into the pre-trained large language model to obtain an updated question-answer pair;

[0161] Scoring the updated question-answer pair to obtain a second score for the updated question-answer pair;

[0162] If the second score satisfies the score condition, the updated question-answer pair is determined as the target question-answer pair.

[0163] In some embodiments, the knowledge practice device further comprises:

[0164] A first acquisition module is used to acquire a sample knowledge data set and construct a sample prompt word set based on the sample knowledge data set;

[0165] A second determination module is used to input the sample prompt word set into a preset instruction language model to obtain an instruction fine-tuning data set;

[0166] The training module is used to adopt a preset model fine-tuning method and use the instruction fine-tuning data set to fine-tune the preset large language model to obtain a pre-trained large language model.

[0167] In some embodiments, the multimodal knowledge data includes document knowledge data, audio knowledge data, and video knowledge data; the second conversion module is specifically used to:

[0168] Recognize the document knowledge data using a preset optical character recognition model to obtain first knowledge data;

[0169] Extracting audio data from video knowledge data;

[0170] Identify the audio knowledge data and the audio data to obtain second knowledge data;

[0171] The textual knowledge data is determined according to the first knowledge data and the second knowledge data.

[0172] In some embodiments, the preset optical character recognition model includes a first character recognition model, a second character recognition model, and a third character recognition model connected in parallel, wherein the first character recognition model, the second character recognition model, and the third character recognition model are respectively models obtained by training the character recognition model through samples in text form, samples in table form, and samples in image format; the conversion module is specifically further used for:

[0173] Convert document knowledge data into knowledge data in a preset format;

[0174] performing version analysis on the knowledge data in a preset format to identify at least one of text data in a text format, table data in a table format, and image data in an image format;

[0175] Using a first character recognition model to recognize text data to obtain first recognition data; and / or,

[0176] Using a second character recognition model to recognize the table data to obtain second recognition data; and / or,

[0177] Using a third character recognition model to recognize the image data to obtain third recognition data; and / or,

[0178] The first knowledge data is determined based on the first identification data, and / or the second identification data, and / or the third identification data.

[0179] In some embodiments, the second character recognition model includes a plurality of second character recognition sub-models; and the conversion module is further configured to:

[0180] Extract feature vector data corresponding to the table data;

[0181] Using the feature vector data as input of each second character recognition sub-model, and obtaining a plurality of recognition results output by each second character recognition sub-model;

[0182] A target recognition result is selected from multiple recognition results by weighted voting, and the target recognition result is used as the second recognition data.

[0183] In some embodiments, the knowledge practice device further comprises:

[0184] A third acquisition module is used to determine the category practice result data of each category knowledge point in each round of knowledge practice based on the practice result data of each round of knowledge practice;

[0185] A statistics module is used to count the comprehensive practice result data of each category of knowledge points according to the category practice result data of each category of knowledge points in each round of knowledge practice;

[0186] The first generating module is used to generate a radar chart of each category of knowledge points according to the comprehensive exercise result data of each category of knowledge points.

[0187] Based on the same inventive concept, the embodiment of the present application also provides a computer device, which may be a server or a terminal device, and includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps of the above-mentioned knowledge practice method when executing the computer program. Thereby, various functions are realized, such as:

[0188] In response to the knowledge practice instruction, obtaining a knowledge practice topic from the knowledge practice instruction;

[0189] Extracting knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer;

[0190] receiving an initial response answer set of the first knowledge question set, and determining exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set;

[0191] Based on the practice result data, questions are re-drawn from the knowledge question bank to obtain a second knowledge question set, and the next round of knowledge practice is performed on the second knowledge question set until the practice end condition is met.

[0192] The knowledge practice method in this embodiment first obtains the knowledge practice topic from the knowledge practice instruction, and extracts multiple knowledge questions from the knowledge question bank that matches the knowledge practice topic, which can ensure that the knowledge questions for the knowledge practice are related to the knowledge practice topic, so that the knowledge questions for the knowledge practice are targeted; then, the initial response answer set of the first knowledge question set is received, and based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, the practice result data of a round of knowledge practice is determined, and based on the practice result data, questions are re-drawn from the knowledge question bank to obtain the second knowledge question set, thereby realizing dynamic and cyclic extraction of knowledge questions combined with the practice situation, realizing real-time dynamic question extraction, ensuring that the re-extracted knowledge question set can accurately reflect the knowledge practice situation, and performing multiple rounds of practice on the dynamically extracted knowledge question set, thereby making the knowledge practice more targeted and greatly improving the knowledge practice effect.

[0193] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0194] In one embodiment, the computer device is a terminal device, for example, its internal structure diagram can be as follows: Fig.10As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a knowledge practice method is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.

[0195] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a partial structure related to the present application scheme, and does not constitute a limitation on the computer device to which the present application scheme is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0196] Based on the same inventive concept, an embodiment of the present application also provides a computer-readable storage medium, which may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0197] Since the computer program stored in the computer-readable storage medium can execute any of the knowledge practice methods provided in the embodiments of the present application, the beneficial effects that can be achieved by any of the knowledge practice methods provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0198] Based on the same inventive concept, the embodiment of the present application further provides a computer program product or a computer program, which includes a computer instruction, and the computer instruction is stored in a computer-readable storage medium. The processor of the computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method provided in various optional implementations in the above embodiments.

[0199] It should be noted that the object data (including but not limited to user device information, user personal information, etc.) and conversation data involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions. A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0200] Any reference to the memory, database or other media used in the embodiments provided in the present application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (Re RAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0201] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0202] In the above-mentioned knowledge practice device, computer-readable storage medium, computer equipment, and computer program product embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process and beneficial effects of the above-described knowledge practice device, computer-readable storage medium, computer program product, computer equipment, and its corresponding units can refer to the description of the knowledge practice method in the above embodiment, and will not be repeated here.

[0203] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0204] The above is a detailed introduction to a knowledge practice method, device, computer equipment, computer-readable storage medium and computer program product provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A knowledge practice method, characterized in that: The method comprises: In response to the knowledge practice instruction, obtaining a knowledge practice topic from the knowledge practice instruction; Extracting knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer; receiving an initial response answer set of the first knowledge question set, and determining exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set; Based on the practice result data, questions are re-drawn from the knowledge question bank to obtain a second knowledge question set, and the next round of knowledge practice is performed on the second knowledge question set until the practice end condition is met.

2. The method according to claim 1, characterized in that Each knowledge question in the knowledge question bank corresponds to a category knowledge point, and re-selecting questions from the knowledge question bank based on the exercise result data to obtain a second knowledge question set includes: Determine a total category knowledge point set corresponding to all knowledge questions in the knowledge question bank, and a first category knowledge point set corresponding to all knowledge questions in the first knowledge question set; Determine a target category knowledge point set according to the total category knowledge point set, the first category knowledge point set and the exercise result data; Based on the practice result data and the target category knowledge point set, at least one knowledge question is re-extracted from the knowledge question bank to obtain a second knowledge question set.

3. The method according to claim 2, characterized in that The step of determining a target category knowledge point set according to the total category knowledge point set, the first category knowledge point set and the exercise result data comprises: When the total category knowledge point set is the same as the first category knowledge point set, determining the total category knowledge point set as the target category knowledge point set; When the total category knowledge point set is different from the first category knowledge point set, a target category knowledge point set is determined from the total category knowledge point set according to the exercise result data.

4. The method according to claim 3, characterized in that The step of determining a target category knowledge point set from the total category knowledge point set according to the exercise result data comprises: Determine a difference between the total category knowledge point set and the first category knowledge point set to obtain a second category knowledge point set; Determining a target first category knowledge point set from the first category knowledge point set according to the exercise result data; The target first category knowledge point set and the second category knowledge point set are determined as the target category knowledge point set.

5. The method according to claim 2, characterized in that Based on the exercise result data and the target category knowledge point set, at least one knowledge question is re-extracted from the knowledge question bank to obtain a second knowledge question set, including: Determining the extraction weight of knowledge questions of each category knowledge point in the target category knowledge point set according to the exercise result data and the target category knowledge point set; According to the extraction weight, at least one knowledge question is re-extracted from the knowledge question bank to obtain the second knowledge question set.

6. The method according to claim 1, characterized in that The knowledge practice instruction includes a target practice mode; the knowledge questions and reference answers in the knowledge question bank are the default practice mode; Before receiving the initial response answer set of the first knowledge question set and determining the exercise result data of a round of knowledge exercise based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, the method further includes: When the target practice mode is different from the default practice mode, the first knowledge question set of the default practice mode is converted into a first knowledge question set consistent with the target practice mode.

7. The method according to claim 1, characterized in that The knowledge practice instruction includes a target practice mode; the receiving of the initial response answer set of the first knowledge question set, and determining the practice result data of a round of knowledge practice based on the initial response answer set and the reference answer corresponding to each knowledge question in the first knowledge question set, including: when the response answers in the initial response answer set are the target practice mode, and the target practice mode is different from the default practice mode corresponding to the knowledge question bank, converting the response answers in the initial response answer set into an initial response answer set matching the default practice mode; The exercise result data of a round of knowledge exercise is determined according to the initial response answer set consistent with the default exercise mode and the reference answers corresponding to the knowledge questions in the first knowledge question set.

8. The method according to claim 1, characterized in that Before extracting knowledge questions from the knowledge question bank matching the knowledge practice theme to obtain a first knowledge question set, the method further includes: Acquire multimodal knowledge data, and convert the multimodal knowledge data into textual knowledge data in text form; Based on the text knowledge data, a pre-trained large language model is used for analysis to generate the knowledge question bank; and / or, Based on a preset interface, obtaining target knowledge data corresponding to the knowledge practice topic; A knowledge question base is constructed based on the target knowledge data to obtain the knowledge question base.

9. The method according to claim 8, characterized in that The method of analyzing the text knowledge data using a pre-trained large language model to generate the knowledge question bank includes: Based on the text knowledge data, construct a target prompt word; The target prompt word is input into the pre-trained large language model for analysis to obtain the knowledge question bank.

10. The method according to claim 9, characterized in that The step of inputting the target prompt word into the pre-trained large language model for analysis to obtain the knowledge question bank includes: Inputting the target prompt word into the pre-trained large language model to obtain an initial question-answer pair output by the pre-trained large language model, wherein the initial question-answer pair includes an initial question and an initial answer; Scoring the initial question-answer pair to obtain a first score for the initial question-answer pair; Determine, according to the first score and the initial question-answer pair, a target question-answer pair corresponding to the initial question pair, use the initial question in the target question-answer pair as the knowledge question, and use the initial answer in the target question-answer pair as a reference answer corresponding to the knowledge question; The knowledge question bank is generated based on all the acquired knowledge questions and the reference answers corresponding to the knowledge questions.

11. The method according to claim 10, characterized in that The step of determining a target question-answer pair corresponding to the initial question pair according to the first score and the initial question-answer pair includes: If the first score satisfies the score condition, determining the initial question-answer pair as the target question-answer pair; If the first score does not satisfy the score condition, retrieving knowledge data according to the initial question in the initial question-answer pair to obtain updated knowledge data; constructing an updated target prompt word based on the updated knowledge data, and inputting the updated target prompt word into the pre-trained large language model to obtain an updated question-answer pair; Scoring the updated question-answer pair to obtain a second score for the updated question-answer pair; If the second score satisfies the score condition, the updated question-answer pair is determined as the target question-answer pair.

12. The method according to claim 10 or 11, characterized in that The method further comprises: Acquire a sample knowledge data set, and construct a sample prompt word set based on the sample knowledge data set; Inputting the sample prompt word set into a preset instruction language model to obtain an instruction fine-tuning data set; A preset model fine-tuning method is adopted, and the instruction fine-tuning data set is used to perform fine-tuning training on the preset large language model to obtain the pre-trained large language model.

13. The method according to claim 8, characterized in that The multimodal knowledge data includes document knowledge data, audio knowledge data and video knowledge data; the step of acquiring the multimodal knowledge data and converting the multimodal knowledge data into textual knowledge data in text form includes: Recognize the document knowledge data using a preset optical character recognition model to obtain first knowledge data; Extracting audio data from the video knowledge data; Identify the audio knowledge data and the audio data to obtain second knowledge data; The text knowledge data is determined according to the first knowledge data and the second knowledge data.

14. The method according to claim 13, characterized in that The preset optical character recognition model includes a first character recognition model, a second character recognition model and a third character recognition model connected in parallel, wherein the first character recognition model, the second character recognition model and the third character recognition model are respectively models obtained by training character recognition models with samples in text form, samples in table form and samples in image format; the document knowledge data is recognized by using the preset optical character recognition model to obtain the first knowledge data, including: Converting the document knowledge data into knowledge data in a preset format; Performing version analysis on the knowledge data in the preset format to identify at least one of text data in a text format, table data in a table format, and image data in an image format; Using the first character recognition model to recognize the text data to obtain first recognition data; and / or, using the second character recognition model to recognize the table data to obtain second recognition data; and / or, Using the third character recognition model to recognize the image data to obtain third recognition data; and / or, The first knowledge data is determined according to the first identification data, and / or the second identification data, and / or the third identification data.

15. The method according to claim 14, characterized in that The second character recognition model includes a plurality of second character recognition sub-models; the second character recognition model is used to recognize the table data to obtain second recognition data, including: Extracting feature vector data corresponding to the table data; Using the feature vector data as input of each of the second character recognition sub-models, and obtaining a plurality of recognition results output by each of the second character recognition sub-models; A target recognition result is selected from the multiple recognition results by adopting a weighted voting method, and the target recognition result is used as the second recognition data.

16. The method according to claim 2, characterized in that After re-drawing questions from the knowledge question bank based on the exercise result data to obtain a second knowledge question set, and performing a next round of knowledge exercises on the second knowledge question set until an exercise end condition is met, the method further includes: Based on the practice result data of each round of knowledge practice, determine the category practice result data of each category knowledge point in each round of knowledge practice; According to the category practice result data of each category of knowledge points in each round of knowledge practice, the comprehensive practice result data of each category of knowledge points are counted; A radar chart of each category of knowledge point is generated according to the comprehensive practice result data of each category of knowledge point.

17. A knowledge training device, characterized in that: The device comprises: An acquisition module, configured to acquire a knowledge practice topic from the knowledge practice instruction in response to the knowledge practice instruction; A question extraction module is used to extract knowledge questions from a knowledge question bank matching the knowledge exercise theme to obtain a first knowledge question set, wherein each knowledge question corresponds to a reference answer; a determination module, configured to receive an initial response answer set of the first knowledge question set, and determine exercise result data of a round of knowledge exercise based on the initial response answer set and a reference answer corresponding to each knowledge question in the first knowledge question set; The practice module is used to re-select questions from the knowledge question bank based on the practice result data to obtain a second knowledge question set, and perform the next round of knowledge practice on the second knowledge question set until the practice end condition is met.

18. A computer device, characterized in that: It comprises a processor and a memory, wherein the memory stores a plurality of instructions; the processor loads instructions from the memory to execute the steps of the knowledge practice method as claimed in any one of claims 1 to 16.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the steps of the knowledge practice method according to any one of claims 1 to 14.

20. A computer software, characterized in that The computer software includes a computer program, and the computer program is executed by a processor to perform the steps of the knowledge practice method according to any one of claims 1 to 16.

21. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the knowledge practice method according to any one of claims 1 to 16 are implemented.

Images