Prior knowledge point prediction method, device, equipment, medium and program product

Abstract

The specification provides a prerequisite knowledge point prediction method, device, equipment, medium and program product. The method comprises: selecting prerequisite knowledge points of a target knowledge point from a similar knowledge point set, obtaining K candidate prerequisite knowledge points, analyzing a knowledge point chain with a prerequisite relationship therefrom, deleting knowledge points other than the last knowledge point in the knowledge point chain to obtain reference prerequisite knowledge points; obtaining an extended knowledge point set, and if there is a knowledge point ring therein, reducing the knowledge points in the extended knowledge point set based on the knowledge point ring; and determining target prerequisite knowledge points of the target knowledge point based on the reference prerequisite knowledge points in the extended knowledge point set after the reduction operation. The multi-stage cooperative optimization mechanism of the embodiment of the specification not only enhances the comprehensiveness of knowledge point recall, but also improves the accuracy of prerequisite knowledge point prediction through chain simplification and loop elimination strategies.

Description

Technical Field

[0001] This specification relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, medium, and program product for predicting prior knowledge points. Background Technology

[0002] The orderly learning of knowledge points is crucial for improving learning effectiveness. Therefore, in scenarios such as personalized learning recommendations, online education, and corporate training, it is usually necessary to understand the prerequisite knowledge points of each knowledge point in order to support the implementation of subsequent learning path generation, knowledge point recommendation and other related functions.

[0003] In recent years, with the development of machine learning technology, the prediction of prior knowledge points has gradually been completed automatically by models. However, in practical applications, it has been found that the accuracy of the prediction results output by the models is relatively low. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this specification provides a method, apparatus, equipment, medium, and program product for predicting prior knowledge points.

[0005] According to a first aspect of the embodiments of this specification, a method for predicting prior knowledge points is provided, the method comprising:

[0006] From the set of similar knowledge points, select the prerequisite knowledge points for the target knowledge point to obtain K candidate prerequisite knowledge points, where K is any positive integer; the set of similar knowledge points includes knowledge points in the knowledge point library that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the basic knowledge point set; wherein, the initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points for the target knowledge point; the basic knowledge point set includes the predicted basic knowledge points that need to be mastered to master the target knowledge point;

[0007] If K is greater than 1, then analyze the knowledge point chains with prerequisite relationships from the K candidate prerequisite knowledge points. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the prerequisite relationship. Based on the knowledge point chain, delete the knowledge points in the knowledge point chain except for the last knowledge point in the knowledge point chain from the K candidate prerequisite knowledge points to obtain the reference prerequisite knowledge points of the target knowledge point.

[0008] Obtain an extended knowledge point set, and check whether there is a knowledge point cycle with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point cycle exists, then based on the knowledge point cycle, delete the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set, so as to delete the knowledge points with incorrect prerequisite relationships based on the knowledge point cycle, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point cycle.

[0009] Based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion, the target prerequisite knowledge points of the target knowledge point are determined.

[0010] In some possible implementations, the step of selecting the prerequisite knowledge points of the target knowledge point from the set of similar knowledge points to obtain K candidate prerequisite knowledge points includes:

[0011] Select N knowledge points from the set of similar knowledge points, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are the pre-knowledge points of the target knowledge points;

[0012] When the L input knowledge points are determined to be prerequisite knowledge points for the target knowledge point, if there are still unselected knowledge points in the set of similar knowledge points, then M knowledge points are selected from the unselected knowledge points in the set of similar knowledge points, and these M knowledge points and the L knowledge points are taken as the current knowledge points. The process returns to the step of inputting the current knowledge points to the pre-analysis agent to determine whether the M knowledge points are prerequisite knowledge points for the target knowledge point. Wherein, when at least one of the M knowledge points is more suitable as a prerequisite knowledge point for the target knowledge point than at least one of the L knowledge points, at least one of the L knowledge points is determined not to be a prerequisite knowledge point for the target knowledge point, where M is an integer less than or equal to N, and L is an integer greater than or equal to 1.

[0013] Based on the prior knowledge points determined by the prior analysis agent, the K candidate prior knowledge points are obtained.

[0014] In some possible implementations, obtaining the K candidate prior knowledge points based on the prior knowledge points determined by the prior analysis agent includes:

[0015] The prerequisite knowledge points determined by the preceding analysis agent are selected as candidate prerequisite knowledge points, resulting in the K candidate prerequisite knowledge points; or...

[0016] The prerequisite knowledge points determined by the prerequisite analysis agent are input into the prerequisite verification agent to verify whether the input knowledge points are prerequisite knowledge points of the target knowledge points. The prerequisite verification agent and the prerequisite analysis agent have different judgment dimensions for prerequisite knowledge points. The prerequisite knowledge points that pass the verification are used as candidate prerequisite knowledge points to obtain the K candidate prerequisite knowledge points.

[0017] In some possible implementations, the step of analyzing the chain of knowledge points with prior relationships from the K candidate prior knowledge points includes:

[0018] Learning paths are generated for the K candidate prerequisite knowledge points to obtain at least one knowledge point learning path. The at least one knowledge point learning path contains all K candidate prerequisite knowledge points without repetition, and the knowledge points in each knowledge point learning path are arranged in the order of learning.

[0019] Among them, the knowledge point learning path that includes at least two candidate prerequisite knowledge points is the knowledge point chain with a prerequisite relationship.

[0020] In some possible implementations, the step of reducing the number of knowledge points in each reference prerequisite knowledge point and / or each reference prerequisite knowledge point's prerequisite knowledge point from the extended knowledge point set based on the knowledge point ring includes:

[0021] When the knowledge point loop consists of two knowledge points, the knowledge point with a more basic cognitive level is identified from the two knowledge points, and the other knowledge point is deleted from the preceding knowledge points of the more basic cognitive level knowledge point.

[0022] When the knowledge point loop includes two or more knowledge points, the degree of prerequisite dependency corresponding to each prerequisite relationship in the knowledge point loop is analyzed. The degree of prerequisite dependency is used to represent the necessity of mastering the prerequisite knowledge points before learning the subsequent knowledge points. Prerequisite relationships that do not meet the prerequisite dependency condition are identified as breakpoints. From the extended knowledge point set, the prerequisite knowledge points corresponding to the breakpoints are deleted from the prerequisite knowledge points of their subsequent knowledge points. And / or, the knowledge points in the knowledge point loop are sorted according to the learning order to obtain the target learning path. Reference prerequisite knowledge points and / or the prerequisite knowledge points of each reference prerequisite knowledge point that do not conform to the target learning path are deleted from the prerequisite knowledge points of the corresponding knowledge points.

[0023] In some possible implementations, determining the target prerequisite knowledge points for the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion includes:

[0024] The reference prerequisite knowledge points in the expanded knowledge point set after the deletion operation are input into the prerequisite evaluation agent to determine whether the input knowledge points are prerequisite knowledge points of the target knowledge point; wherein, the model used by the prerequisite evaluation agent is different from the model used to select candidate prerequisite knowledge points;

[0025] Delete the reference prior knowledge points that are not prior knowledge points as determined by the prior evaluation agent, and retain the reference prior knowledge points that are determined to be prior knowledge points as the target prior knowledge points of the target knowledge point.

[0026] In some possible implementations, the process of generating the set of similar knowledge points includes:

[0027] For the target knowledge point, query the knowledge point database for knowledge points that meet the similarity conditions to the target knowledge point, and determine the score of the queried knowledge point based on the similarity corresponding to the queried knowledge point;

[0028] For each knowledge point in the initial set of prerequisite knowledge points, query the knowledge point database for knowledge points that meet the similarity conditions with the knowledge point, and multiply the similarity of the queried knowledge point by the first weight coefficient to determine the score of the queried knowledge point.

[0029] For each knowledge point in the set of basic knowledge points, query the knowledge point database for knowledge points that meet the similarity conditions with the knowledge point, and determine the similarity and second weight coefficient of the queried knowledge points as the score of each queried knowledge point. The first weight coefficient and the second weight coefficient are both greater than 1, and the first weight coefficient is greater than the second weight coefficient.

[0030] Knowledge points whose scores exceed a preset threshold among the retrieved knowledge points are identified as knowledge points in the set of similar knowledge points.

[0031] In some possible implementations, the step of querying knowledge points from the knowledge point database that satisfy similarity conditions to the target knowledge point, and determining the score of the queried knowledge points based on their similarity, includes:

[0032] Based on the name of the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the name of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0033] Based on the definition of the target knowledge point, query the knowledge point database for knowledge points whose definitions meet the similarity conditions with the definition of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0034] Based on the concepts involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the concepts involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0035] Based on the professional terms involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the professional terms involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0036] The concepts and professional terms involved in the target knowledge points are obtained through semantic parsing based on the names and definitions of the target knowledge points; each concept and professional term has a corresponding importance parameter, which is used to represent the importance of the corresponding concept or professional term in the target knowledge point; the third weight coefficient is determined based on the importance parameter, and the third weight coefficient is positively correlated with the importance parameter and is greater than 1.

[0037] According to a second aspect of the embodiments of this specification, a pre-knowledge point prediction device is provided, comprising:

[0038] A candidate selection unit is used to select prerequisite knowledge points for the target knowledge point from a set of similar knowledge points, resulting in K candidate prerequisite knowledge points, where K is any positive integer. The set of similar knowledge points includes knowledge points in the knowledge point library that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the basic set of knowledge points. The initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points for the target knowledge point. The basic set of knowledge points includes the predicted basic knowledge points that need to be mastered to master the target knowledge point.

[0039] A direct pre-filtering unit is used to analyze knowledge point chains with pre-relationships from the K candidate pre-relationship knowledge points if K is greater than 1. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the pre-relationship. Based on the knowledge point chain, knowledge points other than the last knowledge point in the knowledge point chain are deleted from the K candidate pre-relationship knowledge points to obtain the reference pre-relationship knowledge points of the target knowledge point.

[0040] The cyclic decomposition unit is used to obtain an extended knowledge point set and check whether there is a knowledge point loop with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point loop exists, the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set are deleted based on the knowledge point loop, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point loop.

[0041] The target determination unit is used to determine the target prerequisite knowledge points of the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion.

[0042] In some possible implementations, the candidate selection unit is specifically used for:

[0043] Select N knowledge points from the set of similar knowledge points, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are the pre-knowledge points of the target knowledge points;

[0044] When the L input knowledge points are determined to be prerequisite knowledge points for the target knowledge point, if there are still unselected knowledge points in the set of similar knowledge points, then M knowledge points are selected from the unselected knowledge points in the set of similar knowledge points, and these M knowledge points and the L knowledge points are taken as the current knowledge points. The process returns to the step of inputting the current knowledge points to the pre-analysis agent to determine whether the M knowledge points are prerequisite knowledge points for the target knowledge point. Wherein, when at least one of the M knowledge points is more suitable as a prerequisite knowledge point for the target knowledge point than at least one of the L knowledge points, at least one of the L knowledge points is determined not to be a prerequisite knowledge point for the target knowledge point, where M is an integer less than or equal to N, and L is an integer greater than or equal to 1.

[0045] Based on the prior knowledge points determined by the prior analysis agent, the K candidate prior knowledge points are obtained.

[0046] In some possible implementations, the candidate selection unit is used to determine the prerequisite knowledge points identified by the prerequisite analysis agent as candidate prerequisite knowledge points, thereby obtaining the K candidate prerequisite knowledge points; or,

[0047] The candidate selection unit is used to input the prerequisite knowledge points determined by the prerequisite analysis agent into the prerequisite verification agent to verify whether the input knowledge points are prerequisite knowledge points of the target knowledge points. The prerequisite verification agent and the prerequisite analysis agent have different judgment dimensions for prerequisite knowledge points. The prerequisite knowledge points that pass the verification are used as candidate prerequisite knowledge points to obtain the K candidate prerequisite knowledge points.

[0048] In some possible implementations, the direct pre-screening unit is used to generate learning paths for the K candidate pre-knowledge points to obtain at least one knowledge point learning path, wherein the at least one knowledge point learning path collectively contains the K candidate pre-knowledge points without repetition, and the knowledge points in each knowledge point learning path are arranged in the order of learning.

[0049] Among them, the knowledge point learning path that includes at least two candidate prerequisite knowledge points is the knowledge point chain with a prerequisite relationship.

[0050] In some possible implementations, the cyclic disassembly unit is specifically used for:

[0051] When the knowledge point loop consists of two knowledge points, the knowledge point with a more basic cognitive level is identified from the two knowledge points, and the other knowledge point is deleted from the preceding knowledge points of the more basic cognitive level knowledge point.

[0052] And / or,

[0053] When the knowledge point loop includes two or more knowledge points, the degree of prerequisite dependency corresponding to each prerequisite relationship in the knowledge point loop is analyzed. The degree of prerequisite dependency is used to represent the necessity of mastering the prerequisite knowledge points before learning the subsequent knowledge points. Prerequisite relationships that do not meet the prerequisite dependency condition are identified as breakpoints. From the extended knowledge point set, the prerequisite knowledge points corresponding to the breakpoints are deleted from the prerequisite knowledge points of their subsequent knowledge points. And / or, the knowledge points in the knowledge point loop are sorted according to the learning order to obtain the target learning path. Reference prerequisite knowledge points and / or the prerequisite knowledge points of each reference prerequisite knowledge point that do not conform to the target learning path are deleted from the prerequisite knowledge points of the corresponding knowledge points.

[0054] In some possible implementations, the target determination unit is specifically used to input reference prerequisite knowledge points from the expanded knowledge point set after deletion into a prerequisite evaluation agent to determine whether the input knowledge points are prerequisite knowledge points of the target knowledge point; wherein the model used by the prerequisite evaluation agent is different from the model used to select candidate prerequisite knowledge points; reference prerequisite knowledge points determined by the prerequisite evaluation agent to be non-prerequisite knowledge points are deleted, and reference prerequisite knowledge points determined to be prerequisite knowledge points are retained as target prerequisite knowledge points of the target knowledge point.

[0055] In some possible implementations, the device further includes:

[0056] The knowledge point recall unit is used to query the knowledge point database for the target knowledge point and find knowledge points that meet the similarity conditions of the target knowledge point, and determine the score of the queried knowledge point based on the similarity of the queried knowledge points.

[0057] The knowledge point recall unit is also used to query knowledge points in the knowledge point database that meet the similarity conditions for each knowledge point in the initial set of prerequisite knowledge points, and to determine the score of the queried knowledge point by multiplying the similarity corresponding to the queried knowledge point with the first weight coefficient.

[0058] The knowledge point recall unit is further configured to query knowledge points in the knowledge point database that meet the similarity conditions for each knowledge point in the basic knowledge point set, and determine the similarity and second weight coefficient of the queried knowledge points as the score of each queried knowledge point, wherein the first weight coefficient and the second weight coefficient are both greater than 1, and the first weight coefficient is greater than the second weight coefficient.

[0059] The knowledge point recall unit is also used to identify knowledge points whose scores exceed a preset threshold among the various knowledge points queried as knowledge points in the set of similar knowledge points.

[0060] In some possible implementations, the knowledge point recall unit is specifically used for:

[0061] Based on the name of the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the name of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0062] Based on the definition of the target knowledge point, query the knowledge point database for knowledge points whose definitions meet the similarity conditions with the definition of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0063] Based on the concepts involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the concepts involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0064] Based on the professional terms involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the professional terms involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0065] The concepts and professional terms involved in the target knowledge points are obtained through semantic parsing based on the names and definitions of the target knowledge points; each concept and professional term has a corresponding importance parameter, which is used to represent the importance of the corresponding concept or professional term in the target knowledge point; the third weight coefficient is determined based on the importance parameter, and the third weight coefficient is positively correlated with the importance parameter and is greater than 1.

[0066] According to a third aspect of the embodiments of this specification, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any of the above embodiments.

[0067] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the above embodiments.

[0068] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the method as described in any of the above embodiments.

[0069] The technical solutions provided in the embodiments of this specification may include the following beneficial effects:

[0070] In this embodiment, based on the target knowledge point itself, its initial set of prerequisite knowledge points, and the set of basic knowledge points, similar knowledge points are recalled to obtain a set of similar knowledge points, from which prerequisite knowledge points are selected. This multi-source recall mechanism significantly expands the search range of prerequisite knowledge points, thus laying a solid foundation for accurately obtaining candidate prerequisite knowledge points. Based on this, for the initially selected candidate prerequisite knowledge points, the chain of knowledge points with prerequisite relationships is further analyzed. By deleting knowledge points in this chain except for the last one, indirect prerequisite knowledge points are removed, while direct prerequisite knowledge points are retained, further improving the accuracy of prerequisite knowledge points. Subsequently, by constructing an expanded set of knowledge points and detecting possible knowledge point loops with prerequisite relationships, erroneous knowledge points that generate loops are dynamically deleted, achieving self-consistency verification and correction of the prerequisite relationship topology. The above multi-stage collaborative optimization mechanism not only enhances the comprehensiveness of knowledge point recall but also improves the accuracy of prerequisite knowledge point prediction layer by layer through a dual strategy of chain simplification and loop elimination.

[0071] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this specification. Attached Figure Description

[0072] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this specification and, together with the description, serve to explain the principles of this specification.

[0073] Figure 1 This is a flowchart illustrating a prior knowledge point prediction method according to an exemplary embodiment of this specification.

[0074] Figure 2 This specification is a flowchart illustrating a method for generating a set of similar knowledge points according to an exemplary embodiment.

[0075] Figure 3 This is a flowchart illustrating a method for querying similar knowledge points based on a target knowledge point, according to an exemplary embodiment.

[0076] Figure 4 This specification is a flowchart illustrating a method for selecting prerequisite knowledge points from a set of similar knowledge points according to an exemplary embodiment.

[0077] Figure 5 This is a flowchart illustrating a prior knowledge point prediction method according to an exemplary embodiment of this specification.

[0078] Figure 6 This is a hardware structure diagram of a computer device containing the pre-knowledge point prediction device in the embodiments of this specification.

[0079] Figure 7 This is a block diagram illustrating a prior knowledge point prediction device according to an exemplary embodiment of this specification.

[0080] Figure 8 This is a block diagram illustrating another prior knowledge point prediction device according to an exemplary embodiment of this specification. Detailed Implementation

[0081] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this specification as detailed in the appended claims.

[0082] The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of this specification. The singular forms “a,” “the,” and “the” as used in this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0083] It should be understood that although the terms first, second, third, etc., may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this specification, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0084] The orderly learning of knowledge points is crucial for improving learning effectiveness. Therefore, in scenarios such as personalized learning recommendations, online education, and corporate training, it is usually necessary to understand the prerequisite knowledge points of each knowledge point in order to support the implementation of subsequent learning path generation, knowledge point recommendation and other related functions.

[0085] In related technologies, the prediction of prerequisite knowledge points is usually achieved based on a system of preset rules or a simple machine learning model. Preset rules are manually defined learning sequences or dependencies between knowledge points. Limited by expert experience and knowledge coverage, they cannot exhaustively cover all knowledge points, especially for emerging fields, interdisciplinary content, and knowledge points not written into preset rules, often failing to effectively predict their prerequisite knowledge points. Meanwhile, simple machine learning models lack a deep understanding of the semantics of knowledge and cannot accurately grasp the intrinsic connections between knowledge points, resulting in low prediction accuracy.

[0086] Based on this, this specification provides a method for predicting prerequisite knowledge points. Through a multi-stage collaborative optimization mechanism, it not only enhances the comprehensiveness of knowledge point recall, but also improves the accuracy of prerequisite knowledge point prediction layer by layer through a dual strategy of chain simplification and loop elimination.

[0087] The embodiments described in this specification will now be described in detail.

[0088] like Figure 1 As shown, Figure 1 This is a flowchart illustrating a prior knowledge point prediction method according to an exemplary embodiment, comprising the following steps:

[0089] S101. Select the prerequisite knowledge points of the target knowledge point from the set of similar knowledge points to obtain K candidate prerequisite knowledge points.

[0090] The target knowledge point can be any knowledge point. In the embodiments of this specification, the target knowledge point is used to represent the knowledge point whose preceding knowledge points are to be predicted.

[0091] The set of similar knowledge points includes knowledge points in the knowledge point base that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the set of basic knowledge points. The initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points for the target knowledge point. The set of basic knowledge points includes the predicted basic knowledge points that need to be mastered to master the target knowledge point.

[0092] Among them, the prerequisite knowledge points for the target knowledge point refer to the knowledge points that need to be mastered before learning or mastering the target knowledge point. The basic knowledge points for the target knowledge point are used to indicate the basic knowledge required to understand the target knowledge point.

[0093] For example, when the target knowledge point is "solving quadratic equations in one variable (quadratic formula)," the prerequisite knowledge points include: "the standard form of quadratic equations in one variable", "substitution and operation of algebraic expressions", "the concept and calculation of square roots", etc.

[0094] The target knowledge point "Solving quadratic equations (quadratic formula)" includes: for the equation ax² + bx + c = 0 (where a ≠ 0), its solution is: x = [-b ± √(b² - 4ac)] / 2a. The knowledge point "Standard form of quadratic equations" includes: recognizing the form ax² + bx + c = 0, and clarifying that a, b, and c are constant coefficients and a ≠ 0. If the equation form is not recognized or the coefficients are not correctly identified, the quadratic formula cannot be used. Therefore, the knowledge point "Standard form of quadratic equations" constitutes a prerequisite knowledge point for the target knowledge point. The knowledge point "Substitution and operation of algebraic expressions" includes: substituting specific numbers into the positions of a, b, and c in the formula. Since substituting specific numbers into the positions of a, b, and c in the formula is the first step in using the formula, one must be able to determine the values ​​of a, b, and c. Therefore, the knowledge point "Substitution and operation of algebraic expressions" constitutes a prerequisite knowledge point for the target knowledge point. The knowledge point "Concept and Calculation of Square Roots" includes understanding the meaning of √ and being able to perform calculations. Since the core step in calculating the quadratic formula x=[-b±√(b²-4ac)] / 2a is calculating √(b²-4ac), the knowledge point "Concept and Calculation of Square Roots" constitutes a prerequisite knowledge point for the target knowledge point.

[0095] For example, when the target knowledge point is "solving quadratic equations in one variable (quadratic formula)," the basic knowledge points for this target knowledge point include: "basic arithmetic operations", "the concepts of rational and irrational numbers", and "order of operations".

[0096] The knowledge point "Basic Arithmetic Operations" includes basic operations such as addition, subtraction, multiplication, division, and exponentiation. Since the simplification process of the quadratic formula involves a series of arithmetic operations, errors in basic operations will lead to incorrect results. Therefore, "Basic Arithmetic Operations" constitutes the foundation of this target knowledge point. The knowledge point "Concepts of Rational and Irrational Numbers" includes the understanding that when the root is not a perfect square, the root is irrational. This knowledge point helps in understanding the essence of the "solution," knowing that the "solution" can be irrational, and simplifying radical expressions. Therefore, "Basic Arithmetic Operations" constitutes the foundation of this target knowledge point. The knowledge point "Order of Operations" includes the order of operations such as parentheses, exponentiation, multiplication, division, addition, and subtraction. Incorrect order of operations is a major cause of calculation failure when calculating the quadratic formula; therefore, "Order of Operations" constitutes the foundation of this target knowledge point.

[0097] This specification's embodiments take into account that relying solely on machine learning models to predict prior knowledge points is susceptible to biases in training data or limitations in semantic understanding, leading to insufficient accuracy in prediction results. Therefore, this specification's embodiments introduce a retrieval-enhanced generation (RAG)-based recall mechanism. By recalling possible prior knowledge points from the knowledge point base and then identifying the true prior knowledge points, this effectively compensates for the model's shortcomings in knowledge coverage and other aspects, significantly improving the accuracy and interpretability of prior knowledge point determination.

[0098] To retrieve more comprehensive knowledge points from the knowledge base (i.e., without omitting any true prerequisite knowledge points), this specification not only retrieves knowledge points from the knowledge base based on the target knowledge point, but also further expands the retrieval by combining the predicted prerequisite and basic knowledge points of the target knowledge point. The retrieved knowledge points are those similar to the three sources mentioned above (i.e., target knowledge point, prerequisite knowledge point, and basic knowledge point). Therefore, these retrieval results are referred to as the similar knowledge point set.

[0099] The following examples illustrate how to generate a set of similar knowledge points:

[0100] In some possible implementations, the process of generating a set of similar knowledge points, such as Figure 2 As shown, the process includes the following steps S201 to S204:

[0101] S201. For the target knowledge point, query the knowledge point database for knowledge points that meet the similarity conditions to the target knowledge point, and determine the score of the queried knowledge point based on the similarity corresponding to the queried knowledge point.

[0102] In some embodiments, a knowledge point base is a database used to store knowledge points. In some embodiments, the knowledge point base stores various types of knowledge points from one or more domains. In some embodiments, the knowledge points in the knowledge point base are stored in a structured format. In some embodiments, a knowledge point in the knowledge point base may include its ID, name, definition, and other information. Of course, the knowledge point in the knowledge point base may also include other information, such as the domain to which the knowledge point belongs, the grade level to which it belongs, the subject to which it belongs, the textbook version to which it belongs, the title of the chapter to which it belongs, the chapter number, the title of the section to which it belongs, and the section number, etc. This specification does not limit the information included in the knowledge point base.

[0103] In some possible implementations, the knowledge point base includes the names and definitions of knowledge points. Similar knowledge points can be queried in the knowledge point base based on both the name and definition of the target knowledge point. For a target knowledge point, knowledge points that meet similarity criteria are queried from the knowledge point base, and a score is determined based on the similarity of the queried knowledge points. This includes: based on the name of the target knowledge point, querying the knowledge point base for knowledge points whose names meet similarity criteria, and determining the score of the queried knowledge points based on their similarity; based on the definition of the target knowledge point, querying the knowledge point base for knowledge points whose definitions meet similarity criteria, and determining the score of the queried knowledge points based on their similarity.

[0104] The similarity condition is used to define the criteria for determining whether two objects being compared are similar. This criterion can be configured according to actual application requirements, and the embodiments in this specification do not impose any limitations on it.

[0105] In some embodiments, the similarity condition includes a similarity threshold, which represents the minimum allowed similarity. Two objects being compared are determined to meet the similarity condition only if their similarity is not less than the similarity threshold.

[0106] In some embodiments, the similarity condition includes a preset rank H, which represents the maximum allowed ranking rank. The two objects being compared are considered to satisfy the similarity condition only if their similarity ranks within the top H (including the Hth) of the similarity ranking of all objects in the knowledge point base with respect to the reference object.

[0107] In some embodiments, knowledge point similarity is determined by comparing the similarity of semantic features. Specifically, based on the name of the target knowledge point, querying the knowledge point database for knowledge points whose names satisfy the similarity condition includes: obtaining the semantic features of the target knowledge point's name; and querying the knowledge point database for knowledge points whose name's semantic features satisfy the similarity condition. The similarity score corresponding to the queried knowledge points is the similarity score between the semantic features of their names and the semantic features of the target knowledge point's name.

[0108] This process involves retrieving knowledge points from a knowledge point database that satisfy the similarity criteria to the definition of the target knowledge point, based on the definition of the target knowledge point. This includes: obtaining the semantic features of the target knowledge point's definition; and retrieving knowledge points from the knowledge point database whose semantic features satisfy the similarity criteria to the semantic features of the target knowledge point's definition. The similarity score of the retrieved knowledge points is the similarity score between the semantic features of their definitions and the semantic features of the target knowledge point's definition.

[0109] In some embodiments, the knowledge point base stores the semantic features of the names of each knowledge point as well as the defined semantic features. Therefore, in the above recall process, the semantic features stored in the knowledge point base can be used directly for similarity determination, without the need to calculate the semantic features of the names of each knowledge point in the knowledge point base and the defined semantic features in real time.

[0110] In some embodiments, the score of a queried knowledge point is determined based on the similarity to the queried knowledge point, including: directly using the similarity to the queried knowledge point as the score of the queried knowledge point. Of course, other calculation methods can also be used to calculate the score based on the similarity, and this specification does not limit this.

[0111] In some possible implementations, to recall more comprehensive knowledge points and avoid omitting genuine prerequisite knowledge points, when recalling similar knowledge points for a target knowledge point, recall can be based not only on the name and definition of the target knowledge point, but also on analyzing the concepts and professional terms involved in the target knowledge point, and recalling based on the concepts and professional data involved in the target knowledge point. For example, Figure 3 As shown, step S201 above includes steps S301 to S304:

[0112] S301. Based on the name of the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the name of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points.

[0113] S302. Based on the definition of the target knowledge point, query the knowledge point database for knowledge points whose definitions meet the similarity conditions with the definition of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points.

[0114] S303. Based on the concepts involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the concepts involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0115] In some embodiments, the concepts involved in the target knowledge point are obtained through semantic parsing based on the name and definition of the target knowledge point. In some embodiments, the concepts involved in a knowledge point refer to the core ideas, abstract principles, or key attributes that constitute the inherent logic of that knowledge point. For example, if the target knowledge point is "distinguishing the concept of a set," then the concepts involved in this target knowledge point include "set," etc.

[0116] In some embodiments, the concept corresponds to an importance parameter, which represents the importance of the corresponding concept within the target knowledge point. A third weighting coefficient is determined based on the importance parameter, and the third weighting coefficient is positively correlated with the importance parameter and greater than 1. By assigning weighting coefficients greater than 1 to the knowledge points recalled based on the concepts involved in the target knowledge point, the recall influence of the concepts involved in the target knowledge point can be effectively improved, resulting in more comprehensive recalled knowledge points and increasing the recall rate of true prerequisite knowledge points.

[0117] In some embodiments, the importance parameter can be expressed in a hierarchical form, such as core, important, general, etc.; or it can be expressed in a numerical form, such as 0.6, 0.9, etc. The embodiments in this specification are merely illustrative examples of the representation of the importance parameter and are not intended to limit it. Of course, the importance parameter can also be expressed in other forms.

[0118] For example, when the importance parameter of a concept is core, the third weighting coefficient can be 1.5. For example, when the importance parameter of a concept is important, the third weighting coefficient can be 1.2.

[0119] The implementation method of "searching for knowledge points whose names are similar to the concepts involved in the target knowledge point from the knowledge point database based on the concept involved in the target knowledge point" is the same as the implementation method of "searching for knowledge points whose names are similar to the name of the target knowledge point from the knowledge point database based on the name of the target knowledge point". You can refer to the above content, and will not repeat them here.

[0120] S304. Based on the professional terms involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the professional terms involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0121] The technical terms involved in the target knowledge point are obtained through semantic parsing based on the name and definition of the target knowledge point. In some embodiments, the technical terms involved in the target knowledge point refer to words or phrases that appear in the name and definition of the target knowledge point and are clearly defined and widely used within a certain discipline, industry, field, or profession.

[0122] In some embodiments, the technical term corresponds to an importance parameter, which represents the importance of the corresponding technical term within the target knowledge point. A third weighting coefficient is determined based on the importance parameter, and the third weighting coefficient is positively correlated with the importance parameter and greater than 1. By assigning a weighting coefficient greater than 1 to the knowledge points recalled based on the technical terms involved in the target knowledge point, the recall influence of the technical terms involved in the target knowledge point can be effectively improved, resulting in more comprehensive recalled knowledge points and increasing the recall rate of true prerequisite knowledge points.

[0123] In some embodiments, the importance parameter can be expressed in a hierarchical form, such as core, important, general, etc.; or it can be expressed in a numerical form, such as 0.6, 0.9, etc. The embodiments in this specification are merely illustrative examples of the representation of the importance parameter and are not intended to limit it. Of course, the importance parameter can also be expressed in other forms.

[0124] For example, when the importance parameter of a certain technical term is core, the third weighting coefficient can be 1.5. For example, when the importance parameter of a certain technical term is important, the third weighting coefficient can be 1.2.

[0125] The implementation method of "searching for knowledge points whose names are similar to the professional terms involved in the target knowledge point from the knowledge point database based on the professional terms involved in the target knowledge point" is the same as the implementation method of "searching for knowledge points whose names are similar to the name of the target knowledge point from the knowledge point database based on the name of the target knowledge point". You can refer to the above content, and will not repeat them here.

[0126] S202. For each knowledge point in the initial set of prerequisite knowledge points, query the knowledge point database for knowledge points that meet the similarity conditions, and determine the score of the queried knowledge point based on the product of the similarity corresponding to the queried knowledge point and the first weight coefficient.

[0127] In some embodiments, the knowledge points in the initial set of prerequisite knowledge points are obtained through prediction, and the prediction result includes the name of the knowledge point. Similar knowledge points can be queried in the knowledge point database based on this name. Specifically, for each knowledge point in the initial set of prerequisite knowledge points, querying the knowledge point database for knowledge points that satisfy similarity conditions includes: for each name in the initial set of prerequisite knowledge points, querying the knowledge point database for knowledge points whose names satisfy similarity conditions.

[0128] In some embodiments, other information such as the definition of a knowledge point can be obtained based on its name in the initial set of prerequisite knowledge points. This allows for the querying of similar knowledge points in the knowledge point database based on both the name and definition of each knowledge point in the initial set of prerequisite knowledge points. For each knowledge point in the initial set of prerequisite knowledge points, querying the knowledge point database for knowledge points that satisfy similarity conditions includes: querying the knowledge point database for knowledge points whose names satisfy similarity conditions to the names of the knowledge points in the initial set of prerequisite knowledge points; and querying the knowledge point database for knowledge points whose definitions satisfy similarity conditions to the definitions of the knowledge points in the initial set of prerequisite knowledge points.

[0129] The implementation method of "searching for similar knowledge points in the knowledge point database based on the name and definition of each knowledge point in the initial set of prerequisite knowledge points" is the same as the implementation method of "searching for similar knowledge points in the knowledge point database based on the name and definition of the target knowledge point". You can refer to the above content, and will not repeat it here.

[0130] In some embodiments, the first weight coefficient is greater than 1. This allows for the assignment of higher scores to knowledge points recalled based on the initial set of prerequisite knowledge points, increasing their likelihood of being retained in the set of similar knowledge points. Since the initial set of prerequisite knowledge points is already a strong candidate for the true prerequisite knowledge points of the target knowledge point, assigning a higher weight coefficient to its expanded recall results helps enhance the coverage of potential prerequisite knowledge points, effectively preventing the omission of true prerequisite knowledge points, and thus improving the accuracy of subsequent prerequisite knowledge point prediction results.

[0131] S203. For each knowledge point in the basic knowledge point set, query the knowledge point database for knowledge points that meet the similarity conditions, and determine the similarity and second weight coefficient of the queried knowledge points as the score of the queried knowledge points.

[0132] In some embodiments, the knowledge points in the basic knowledge point set are obtained through prediction, and the prediction result includes the name of the knowledge point. Similar knowledge points can be queried in the knowledge point base based on this name. Specifically, for each knowledge point in the basic knowledge point set, querying the knowledge point base for knowledge points that satisfy similarity conditions includes: for each knowledge point in the basic knowledge point set, querying the knowledge point base for knowledge points whose names satisfy similarity conditions to the names of those knowledge points.

[0133] In some embodiments, other information such as the definition of a knowledge point can be obtained based on its name in the basic knowledge point set. This allows for the querying of similar knowledge points in the knowledge point database based on both the name and definition of each knowledge point in the basic knowledge point set. For each knowledge point in the basic knowledge point set, querying the knowledge point database for knowledge points that satisfy similarity conditions includes: querying the knowledge point database for knowledge points whose names satisfy similarity conditions to the names of the knowledge points in the basic knowledge point set; and querying the knowledge point database for knowledge points whose definitions satisfy similarity conditions to the definitions of the knowledge points in the basic knowledge point set.

[0134] The implementation method of "searching for similar knowledge points in the knowledge point database based on the name and definition of each knowledge point in the basic knowledge point set" is the same as the implementation method of "searching for similar knowledge points in the knowledge point database based on the name and definition of the target knowledge point". You can refer to the above content, and will not be repeated here.

[0135] In some embodiments, the second weighting coefficient is greater than 1. This allows for the assignment of higher scores to knowledge points recalled based on the set of basic knowledge points, increasing their likelihood of remaining in the set of similar knowledge points. Since basic knowledge points reflect the underlying supporting knowledge necessary for mastering the target knowledge points and are often semantically or logically closely related to the true prerequisite knowledge points, assigning a higher weighting coefficient to their extended recall results helps enhance the coverage of potential prerequisite knowledge points, effectively preventing the omission of true prerequisite knowledge points, and thus improving the accuracy of subsequent prerequisite knowledge point prediction results.

[0136] In some embodiments, considering that the relationship between the knowledge points in the initial set of prerequisite knowledge points and the actual prerequisite knowledge points is closer than the relationship between the knowledge points in the set of basic knowledge points and the actual prerequisite knowledge points, the first weight coefficient is set to be greater than the second weight coefficient.

[0137] In some possible implementations, when querying similar knowledge points from the knowledge point base in steps S201 to S203, the entire knowledge point base can be used as the search scope, or the search scope can be limited according to actual teaching logic or cognitive rationality. For example, the search space can be dynamically narrowed through filtering conditions to improve the relevance and educational applicability of the recall results.

[0138] For example, when executing a query, a grade-level constraint can be set to retrieve only knowledge points in the knowledge point database whose grade level is less than or equal to that of the target knowledge point. This restriction is based on the principle of "cognitive development sequence" in pedagogy—students typically do not learn higher-grade knowledge before mastering lower-grade content; therefore, prerequisite knowledge points generally should not appear in higher-grade textbooks. Similarly, the search scope can be further refined by combining data such as subject, textbook version, and semester module, thereby ensuring broad knowledge coverage while effectively eliminating distracting items that do not conform to pedagogical logic, improving the accuracy and practicality of similar knowledge point sets.

[0139] S204. Among the various knowledge points retrieved, those with scores exceeding a preset threshold are identified as knowledge points in the set of similar knowledge points.

[0140] The preset threshold can be any value, such as 0.6, 0.7, etc. The embodiments in this specification do not limit the preset threshold.

[0141] In some embodiments, since the embodiments of this specification use different query methods (that is, recalling similar knowledge points from the knowledge point database based on the target knowledge point, each knowledge point in the initial prerequisite knowledge point set and each knowledge point in the basic knowledge point set respectively), the recalled knowledge points may be duplicated. Therefore, the queried knowledge points can be deduplicated first, and then the knowledge points with scores exceeding the preset threshold can be determined as knowledge points in the similar knowledge point set.

[0142] It should be noted that the embodiments in this specification only use steps S201 to S203 above as examples to illustrate the calculation method of the score for the recalled knowledge points. Of course, other calculation methods can also be used to calculate the score for the recalled knowledge points, and the embodiments in this specification do not limit this. For example, regardless of whether the knowledge points are recalled based on the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, or each knowledge point in the set of basic knowledge points, the similarity corresponding to the knowledge point is directly determined as the score of the knowledge point.

[0143] Another point to note is that this specification only uses step S204 above as an example to illustrate the screening of recall results. In other embodiments, different screening strategies can be adopted. For example, the G knowledge points with the highest scores can be selected as knowledge points in the similar knowledge point set, where G is any positive integer. Alternatively, a combination of both can be used, selecting knowledge points that simultaneously satisfy the condition of "score exceeding a preset threshold" and falling within the "top G" range. The above screening methods can be flexibly configured according to the size of the knowledge point base and the application scenario's requirements for recall accuracy and coverage, thereby ensuring relevance while also considering the diversity and practicality of the candidate set.

[0144] Next, we will provide an example of how to obtain the initial set of prerequisite knowledge points and the set of basic knowledge points:

[0145] In some possible implementations, the initial set of prerequisite knowledge points and the set of basic knowledge points are obtained through prediction by a large language model. The process of generating the initial set of prerequisite knowledge points and the set of basic knowledge points includes: constructing a first prompt word and inputting it into the large language model; this first prompt word instructs the large language model to output the prerequisite knowledge points and basic knowledge points of the target knowledge point based on its name and definition; and obtaining the prerequisite knowledge points and basic knowledge points output by the large language model to obtain the initial set of prerequisite knowledge points and the set of basic knowledge points.

[0146] In some embodiments, constructing a first prompt word includes: obtaining a first prompt word template, which includes: a requirement description related to generating prerequisite and basic knowledge points, and placeholders for filling in the name and definition of the target knowledge point; replacing the corresponding placeholders in the first prompt word template with the name and definition of the target knowledge point to obtain the first prompt word. In some embodiments, the requirement description includes a task purpose description (e.g., analyzing the direct prerequisite and basic knowledge points of the target knowledge point) and a task execution rule description (e.g., prerequisite knowledge should conform to the grade arrangement of the teaching syllabus). The task purpose description instructs the large language model what to do, and the task execution rule description instructs the large language model how to do it.

[0147] In some possible implementations, the concepts and technical terms involved in the target knowledge point are also predicted through a large language model. The initial set of prerequisite knowledge points, the set of basic knowledge points, and the concepts and technical terms involved in the target knowledge point can be generated together in a single large language model call, or they can be generated step-by-step through multiple independent large language model calls. This specification does not limit this approach; it only exemplifies generating them together in a single large language model call. In some embodiments, the generation process of the initial set of prerequisite knowledge points, the set of basic knowledge points, and the concepts and technical terms involved in the target knowledge point includes: constructing a second prompt word and inputting it into the large language model, the second prompt word instructing the large language model to output the prerequisite knowledge points, basic knowledge points, and the concepts and technical terms involved in the target knowledge point based on the name and definition of the target knowledge point; obtaining the prerequisite knowledge points, basic knowledge points, and the concepts and technical terms involved in the target knowledge point output by the large language model, thus obtaining the initial set of prerequisite knowledge points, the set of basic knowledge points, and the concepts and technical terms involved in the target knowledge point.

[0148] In some embodiments, the second prompt word is also used to instruct the large language model to output the importance parameters corresponding to the concepts and technical terms involved in the target knowledge point.

[0149] In some embodiments, constructing a second prompt word includes: obtaining a second prompt word template, which includes: a requirement description related to the concepts and professional terms involved in generating prerequisite knowledge points, basic knowledge points, and target knowledge points, and placeholders for filling in the name and definition of the target knowledge point; replacing the corresponding placeholders in the second prompt word template with the name and definition of the target knowledge point to obtain the second prompt word. In some embodiments, the requirement description includes a task purpose description and a task execution rule description. The task purpose description instructs the large language model what to do, and the task execution rule description instructs the large language model how to do it.

[0150] Next, this specification provides an exemplary description of how to select the prerequisite knowledge points for a target knowledge point from a set of similar knowledge points, such as... Figure 4 As shown, the process of selecting prerequisite knowledge points for a target knowledge point from a set of similar knowledge points may include the following steps:

[0151] S401. Select N knowledge points from the set of similar knowledge points, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are the pre-knowledge points of the target knowledge points.

[0152] Where N is any positive integer. For example, N can be 3, 5, etc. This specification does not limit the value of N in its embodiments. The pre-analysis agent is used to analyze whether the knowledge points in the set of similar knowledge points are pre-knowledge points of the target knowledge point.

[0153] In some possible implementations, the current knowledge point is input into a pre-analysis agent to determine whether the input knowledge point is a pre-knowledge point of the target knowledge point. This includes: inputting the current knowledge point into the pre-analysis agent; the pre-analysis agent constructing a third prompt word based on the current knowledge point and the target knowledge point; this third prompt word instructing the large language model to analyze whether the current knowledge point is a pre-knowledge point of the target knowledge point; the pre-analysis agent inputting the third prompt word into the large language model to obtain the analysis result output by the large language model, which indicates whether the current knowledge point is a pre-knowledge point of the target knowledge point. In some embodiments, the large language model may also output a reason for the judgment of the current knowledge point.

[0154] In some embodiments, the pre-analysis agent stores a preset third prompt word template. This template includes: a requirement description related to determining whether the current knowledge point is a prerequisite knowledge point for the target knowledge point; and placeholders for filling in the target knowledge point and the current knowledge point, respectively. The third prompt word is obtained by replacing the corresponding placeholders in the template with the target knowledge point and the current knowledge point. In some embodiments, the requirement description includes a task purpose description and a task execution rule description. The task purpose description instructs the large language model what to do, and the task execution rule description instructs the large language model how to do it.

[0155] Step S401 selects only N knowledge points from the set of similar knowledge points as the current knowledge point input to the pre-analysis agent for analysis. Compared to inputting the entire set of similar knowledge points into the model for batch analysis, this significantly reduces the cognitive load and semantic interference on the model. Because the input size is controllable, the pre-analysis agent can devote more attention resources to each input knowledge point, performing more refined reasoning, and thus more accurately identifying whether it constitutes a true pre-knowledge point. This strategy effectively avoids ambiguity or misjudgment caused by overloaded input information, especially when dealing with semantically similar but logically unrelated knowledge points, significantly improving the accuracy and reliability of pre-knowledge relationship determination.

[0156] S402. When the L input knowledge points are determined to be prerequisite knowledge points for the target knowledge point, if there are still unselected knowledge points in the set of similar knowledge points, then select M knowledge points from the unselected knowledge points in the set of similar knowledge points and take the M knowledge points and the L knowledge points as the current knowledge points. Return to the step of inputting the current knowledge points to the prerequisite analysis agent to determine whether the M knowledge points are prerequisite knowledge points for the target knowledge point.

[0157] M is an integer less than or equal to N. In some embodiments, M can be equal to N when the number of unselected knowledge points in the set of similar knowledge points is not less than N. When the number of unselected knowledge points in the set of similar knowledge points is less than N, M can be the number of unselected knowledge points in the set of similar knowledge points.

[0158] L is an integer greater than or equal to 1. After executing step S401 above, if all the knowledge points input to the pre-analysis agent are determined not to be pre-knowledge points of the target knowledge point, then continue executing step S401, that is, continue to select N knowledge points from the unselected knowledge points in the similar knowledge point set, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are pre-knowledge points of the target knowledge point. If at least one knowledge point input to the pre-analysis agent is determined to be a pre-knowledge point of the target knowledge point, execute step S402 above.

[0159] In some embodiments, when at least one of the M knowledge points is more suitable as a prerequisite knowledge point for the target knowledge point than at least one of the L knowledge points, at least one of the L knowledge points is determined not to be a prerequisite knowledge point for the target knowledge point.

[0160] In step S402 above, the pre-analysis agent supports dynamic re-evaluation of pre-relationships. When at least one of the newly introduced M knowledge points is more suitable as a pre-relationship of the target knowledge point than some of the original L knowledge points, the pre-analysis agent can identify and correct its previous judgment, reclassifying the relatively unsuitable knowledge points from the original L knowledge points as "non-pre-relationship knowledge points." Thus, when judging newly introduced knowledge points, the pre-analysis agent can also self-correct existing conclusions, effectively avoiding misjudgments caused by insufficient local information and significantly improving the accuracy of pre-relationship knowledge point identification results.

[0161] In some possible implementations, the step of selecting M knowledge points from the unselected knowledge points included in the set of similar knowledge points and using these M knowledge points and L knowledge points as the current knowledge points, and then inputting the current knowledge points into the pre-analysis agent to determine whether the M knowledge points are pre-selected knowledge points of the target knowledge point, includes: selecting M knowledge points from the unselected knowledge points included in the set of similar knowledge points and using these M knowledge points and L knowledge points as the current knowledge points; inputting the current knowledge points into the pre-analysis agent; the pre-analysis agent constructing a fourth prompt word based on the current knowledge points and the target knowledge points; the fourth prompt word being used to instruct the large language model to analyze whether the undetermined knowledge points in the current knowledge points are pre-selected knowledge points of the target knowledge point, and verifying the pre-selected knowledge points that are determined to be pre-selected knowledge points of the target knowledge point; the pre-analysis agent inputting the fourth prompt word into the large language model to obtain the analysis result output by the large language model, which is used to indicate whether the current knowledge point is a pre-selected knowledge point of the target knowledge point.

[0162] In some embodiments, the large language model can also output the reasoning behind the judgment of the current knowledge point.

[0163] In some embodiments, the pre-analysis agent stores a preset fourth prompt word template. This template includes: a requirement description related to determining whether an unidentified knowledge point in the current knowledge point is a prerequisite knowledge point for the target knowledge point, and reviewing the prerequisite knowledge points identified as the target knowledge point; and placeholders for filling in the target knowledge point, the unidentified knowledge point, and the prerequisite knowledge points identified as the target knowledge point, respectively. The fourth prompt word is obtained by replacing the corresponding placeholders in the fourth prompt word template with the target knowledge point, the unidentified knowledge point, and the prerequisite knowledge points identified as the target knowledge point. In some embodiments, the requirement description includes a task purpose description and a task execution rule description.

[0164] It should be noted that the embodiments in this specification are only examples of inputting knowledge points from the set of similar knowledge points into the pre-analysis agent in batches for analysis, to illustrate the method of selecting target knowledge points from the set of similar knowledge points. In another embodiment, all knowledge points in the set of similar knowledge points can be input into the pre-analysis agent together for analysis. The embodiments in this specification do not limit this.

[0165] S403. Based on the prior knowledge points determined by the prior analysis agent, obtain the K candidate prior knowledge points.

[0166] In some embodiments, the prerequisite knowledge points determined by the prerequisite analysis agent are the prerequisite knowledge points determined by the agent last time. For example, the similar knowledge point set includes 12 knowledge points. First, the first 3 knowledge points are input into the prerequisite analysis agent, which identifies prerequisite knowledge points that do not contain the target knowledge point. Then, the 4th to 6th knowledge points are input into the agent, which identifies the 5th knowledge point as a prerequisite knowledge point for the target knowledge point. Next, the 5th knowledge point and the 7th to 9th knowledge points are input into the agent together, which determines that the 5th knowledge point is not a prerequisite knowledge point for the target knowledge point, and the 7th knowledge point is a prerequisite knowledge point for the target knowledge point. Finally, the 7th knowledge point and the 10th to 12th knowledge points are input into the agent together, which identifies the 7th and 12th knowledge points as prerequisite knowledge points for the target knowledge point. Therefore, the prerequisite knowledge points determined by the pre-analysis agent are the 7th and 12th knowledge points, excluding the 5th knowledge point.

[0167] In some possible implementations, the prerequisite knowledge points determined by the prerequisite analysis agent can be directly used as candidate prerequisite knowledge points. Specifically, obtaining K candidate prerequisite knowledge points based on these prerequisite knowledge points includes: identifying the prerequisite knowledge points determined by the prerequisite analysis agent as candidate prerequisite knowledge points, thus obtaining K candidate prerequisite knowledge points. Here, K is any positive integer.

[0168] In some possible implementations, the prerequisite knowledge points determined by the pre-analysis agent can also be verified, and the verified knowledge points are used as candidate prerequisite knowledge points. Specifically, obtaining K candidate prerequisite knowledge points based on the prerequisite knowledge points determined by the pre-analysis agent includes: inputting the prerequisite knowledge points determined by the pre-analysis agent into a pre-verification agent to verify whether the input knowledge points are prerequisite knowledge points of the target knowledge point; wherein the pre-verification agent and the pre-analysis agent have different judgment dimensions for the prerequisite knowledge points; and using the verified prerequisite knowledge points as candidate prerequisite knowledge points to obtain K candidate prerequisite knowledge points.

[0169] In some embodiments, the preceding knowledge points determined by the preceding analysis agent are input into the preceding verification agent to verify whether the input knowledge points are preceding knowledge points of the target knowledge point. This includes: inputting the preceding knowledge points determined by the preceding analysis agent into the preceding verification agent; the preceding verification agent constructing a fifth prompt word based on the preceding knowledge points determined by the preceding analysis agent and the target knowledge point; the fifth prompt word being used to instruct the large language model to verify whether the input preceding knowledge points are true preceding knowledge points of the target knowledge point; and the preceding verification agent inputting the fifth prompt word into the large language model to obtain the analysis result output by the large language model, which is used to indicate whether the input preceding knowledge points are true preceding knowledge points of the target knowledge point.

[0170] In some embodiments, the large language model can also output the reasoning behind the judgments of each prerequisite knowledge point.

[0171] In some embodiments, the pre-verification agent stores a preset fifth prompt word template, which includes: a requirement description related to verifying whether the pre-knowledge point of the input is a true pre-knowledge point of the target knowledge point, and placeholders for filling in the target knowledge point and the pre-knowledge point respectively; replacing the corresponding placeholders in the fifth prompt word template with the target knowledge point and the pre-knowledge point yields the fifth prompt word. In some embodiments, the requirement description includes a task purpose description and a task execution rule description.

[0172] The pre-verification agent and the pre-analysis agent use different judgment dimensions for the pre-existing knowledge points. Therefore, the task execution rule descriptions included in the fifth prompt word template differ from those included in the third and fourth prompt word templates. Considering that too many judgment dimensions would not only lead to overly complex prompt words but also potentially weaken the focus on each judgment dimension, thus affecting overall accuracy, the task is decoupled into two stages, making the judgment results of each stage more accurate.

[0173] In some embodiments, the pre-verification agent is used to perform a more in-depth analysis of the pre-knowledge points determined by the pre-analysis agent.

[0174] To make it easier to understand, the following steps will be explained with specific examples:

[0175] First, prepare basic services:

[0176] A high-performance embedding model service is built using the Infinity framework (a high-performance service framework for deploying embedding models) and the FastAPI framework. Text is input into this embedding model service, which outputs the embedding vector of that text, i.e., the semantic features of the text. The embedding model can be the gte-Qwen2-7B-instruct model (a general-meaning Qwen2-7B-instruction fine-tuning model optimized for text embedding tasks).

[0177] The sglang (Structured Generation Language, a programming framework for structured large language model inference, mainly used to control the large language model to generate output that conforms to a specific syntax or format) framework is used to deploy the large language model service API interface for workflow calls. The large language model can be a QwQ-32B model.

[0178] Deploy the Milvus vector library (an open-source vector database) using Docker (an open-source containerization platform), and deploy a CRUD (Create, Read, Update, Delete) API interface using the FastAPI framework to facilitate operations by the Milvus vector library.

[0179] Then, the knowledge points are stored in the Milvus vector library, resulting in a knowledge point library. When storing knowledge points in the Milvus vector library, they can be organized into structured data, such as:

[0180] {

[0181] "id" (knowledge point identifier): "35037614"

[0182] "knowledge_ponit" (Knowledge Point Name): "Clarifying the Concept of Sets"

[0183] "Definition": "The concept of a set - the clarification of the concept of a set mainly involves determining whether a group of objects can constitute a set. The elements of a set must satisfy the following three properties: determinism (each element is definite), distinctness (each element is unique), and unorderedness (the order of the elements does not affect the set). Through these properties, we can determine which objects can form a set."

[0184] "grade": "Grade 11", #grade

[0185] "subject": "mathematics", #subject

[0186] "version": "People's Education Edition (Version A)", #Textbook Version

[0187] "semester": "Required Course 1", #semester

[0188] "level1_knowledge_point": "Chapter 1 Sets and Common Logical Terms", #The name of the chapter.

[0189] "chapter": 1, #Chapter number

[0190] "level2_knowledge_point": "1.1 The concept of sets", # the name of the section to which it belongs.

[0191] "section": 1.01, #section number

[0192] "embedding": [...], #Embedding (a knowledge point definition)

[0193] "knowledge_point_embedding": [...] #Knowledge point name Embedding

[0194] }

[0195] After organizing the various knowledge points into the structured data shown above, the corresponding API interface is called to store them in the Milvus vector library, thus obtaining the knowledge point library.

[0196] Next, you can select a knowledge point from the knowledge point base as the target knowledge point. For example... Figure 5As shown, the prediction agent is invoked, and the target knowledge point (which can be all the data of the target knowledge point, such as the structured data shown above, or partial data, such as the name and definition of the target knowledge point) is input into the prediction agent. The prediction agent constructs a second prompt word based on the target knowledge point and the second prompt word template, and inputs the second prompt word into the large language model. The large language model outputs the concepts and terms involved in the target knowledge point, the importance parameters corresponding to the concepts and terms, the prerequisite knowledge points of the target knowledge point, and the basic knowledge points of the target knowledge point. Among these, the prerequisite knowledge point output by the large oracle model is the name of that prerequisite knowledge point, and the basic knowledge point output is the name of that basic knowledge point.

[0197] For example, the requirement description in the second prompt word template can be as follows:

[0198] "You are an experienced teaching and research teacher. Please help me analyze the prerequisite knowledge system for the target knowledge points:"

[0199] # The dimensions of analysis are as follows

[0200] 1. Breakdown of core concepts:

[0201] - Key Concepts: What is the most crucial concept in this knowledge point?

[0202] - Key terms: The technical terms (such as formulas, theorems, etc.) involved in this knowledge point.

[0203] 2. Necessary prerequisite knowledge system:

[0204] - Direct prerequisites: What knowledge points must be mastered before understanding the target knowledge point?

[0205] - Foundational Support: What fundamental knowledge will be used to understand the target knowledge points?

[0206] #Important Notes

[0207] 1. Please analyze based on the cognitive level of students in grade {}.

[0208] 2. Prerequisite knowledge should conform to the grade-level arrangement in the teaching syllabus.

[0209] 3. The importance parameter has two possible values: [Core / Important]. Please determine the value based on the importance of the concepts and technical terms.

[0210] 4. The type has two possible values: [Prerequisite / Basic]. Please determine the type based on the relationship between the knowledge points.

[0211] 5. The analysis should highlight the logical progression between knowledge points, and these relationships should be explained in detail in the "Reflection" section.

[0212] Then, based on the target knowledge point and the output of the prediction agent, similar knowledge points are retrieved from the knowledge point base. Specifically:

[0213] First, the embedding model service is invoked to obtain the embedding representations of each concept involved in the target knowledge point, each professional term involved in the target knowledge point, each prerequisite knowledge point name, and each basic knowledge point name. Then, based on these embedding representations, the Milvus query interface is invoked to query the `knowledge_point_embedding` field, returning knowledge points whose similarity exceeds a similarity threshold. Simultaneously, based on the embedding representation of the target knowledge point name, the Milvus query interface is invoked to query the `knowledge_point_embedding` field, returning knowledge points whose similarity exceeds a similarity threshold; and based on the embedding representation defined for the target knowledge point, the Milvus query interface is invoked to query the `embedding` field, returning knowledge points whose similarity exceeds a similarity threshold.

[0214] When querying based on the name and definition of a target knowledge point, the search scope can be fixed using a Milvus filter statement (of course, if these fields do not exist, they can be ignored). The Milvus filter statement can be shown below:

[0215] filter_where = f"(metadata['grade']<= {search_terms['grade']}) and \# The grade level of the knowledge point being queried must be less than or equal to the grade level of the knowledge point being queried.

[0216] (metadata['chapter']<= {search_terms['chapter']}) and \ # The chapter of the knowledge point being queried must be less than or equal to the chapter of the knowledge point being queried.

[0217] (metadata['section']<= {search_terms['section']}) and \ # The section of the searched knowledge point must be less than or equal to the section of the knowledge point entered in the search.

[0218] (metadata['knowledge_point'] != '{search_terms['knowledge_point']}')"# The knowledge points queried cannot be equal to the knowledge points entered in the query.

[0219] After obtaining all query results, duplicates can be removed, and then weighted for calculation. For example: for knowledge points retrieved based on core concepts, core professional terms, or prerequisite knowledge points, the similarity score is multiplied by 1.5 to obtain the score for that knowledge point. For knowledge points retrieved based on important concepts, important core terms, or basic knowledge points, the similarity score is multiplied by 1.2 to obtain the score for that knowledge point. For knowledge points retrieved based on the name or definition of the target knowledge point, the similarity score is multiplied by 1.0 to obtain the score for that knowledge point.

[0220] After calculating the score for each knowledge point, the knowledge points can be sorted in descending order of score, and knowledge points with scores less than 0.6 can be filtered out. The top 12 knowledge points are then used to generate a set of similar knowledge points.

[0221] Then, select 3 knowledge points from the set of similar knowledge points, call the first pre-analysis agent (where the pre-analysis agent includes the first pre-analysis agent and the iterative pre-analysis agent) to analyze these 3 knowledge points, determine whether each of these 3 knowledge points is a pre-knowledge point of the target knowledge point, and output the analysis results and the reason for the judgment. If no knowledge point is identified as a prerequisite for the target knowledge point by the initial pre-analysis agent (i.e., none of the knowledge points input into the initial pre-analysis agent are prerequisites for the target knowledge point), then three knowledge points are selected from the set of similar knowledge points that were not selected, and the initial pre-analysis agent is invoked to analyze these three knowledge points. If a knowledge point is identified as a prerequisite for the target knowledge point by the initial pre-analysis agent, then three more knowledge points are selected from the set of similar knowledge points that were not selected, and these are added to the knowledge points already identified as prerequisites for the target knowledge point. The iterative pre-analysis agent is invoked to analyze these knowledge points, determining whether each knowledge point is a prerequisite for the target knowledge point. The knowledge points identified as prerequisites for the target knowledge point are then reviewed. If any unsuitable knowledge points are found, they need to be deleted, and the reasons given. If the output analysis result indicates that none of the input knowledge points are prerequisites for the target knowledge point, then the initial pre-analysis agent is invoked again in the next iteration. In other words, if there is no knowledge point that is identified as a prerequisite knowledge point for the target knowledge point, the initial prerequisite analysis agent is invoked; if there is a knowledge point that is identified as a prerequisite knowledge point for the target knowledge point, the iterative prerequisite analysis agent is invoked.

[0222] The initial pre-analysis agent analyzes knowledge points by calling a large language model. Specifically, the initial pre-analysis agent stores a preset third prompt word template. Based on this template, knowledge points selected from a set of similar knowledge points, and the target knowledge point, the agent constructs a third prompt word. This third prompt word is then input into the large language model to obtain the analysis result output by the model. This result indicates whether the knowledge point selected from the set of similar knowledge points is a pre-knowledge point of the target knowledge point.

[0223] For example, the requirement description in the third prompt word template can be as follows:

[0224] The analysis dimensions are as follows:

[0225] 1. Necessary dependency:

[0226] - Does the definition or application of the target knowledge point directly use the prerequisite knowledge point?

[0227] - If this prerequisite knowledge point is removed, can the target knowledge point still be fully understood and applied?

[0228] Note: The order in which knowledge is learned does not necessarily establish a prerequisite knowledge relationship.

[0229] 2. Concept Construction:

[0230] - Are the core concepts of the target knowledge point directly built upon the prerequisite knowledge points?

[0231] - Do the prerequisite knowledge points include the basic concepts necessary for understanding the target knowledge point?

[0232] This conceptual dependency must be clearly reflected in the definition of the target knowledge point.

[0233] 3. Practical Applications:

[0234] - When applying the target knowledge points, is it necessary to use the prerequisite knowledge points?

[0235] These methods or skills must be an indispensable part of applying the target knowledge points.

[0236] - Rather than a dispensable auxiliary tool

[0237] 4. Direct association:

[0238] - Are there direct conceptual references or methodological applications between the prerequisite knowledge points and the target knowledge points?

[0239] - Avoid misclassifying indirectly related knowledge points or knowledge points that only appear earlier in the teaching sequence as prerequisite knowledge points.

[0240] - Focus on the substantive dependencies between concepts

[0241] Precautions:

[0242] - The judgment is based strictly on the definition of the target knowledge points, rather than on the traditional teaching sequence.

[0243] - Prerequisite knowledge points must substantially help in mastering the target knowledge point; without mastering the prerequisite knowledge points, it is impossible to master the target knowledge point.

[0244] - Simply having a specific learning order is not a sufficient condition for recognizing prerequisite knowledge points.

[0245] - Remain objective and explain your analysis in detail.

[0246] - The analysis needs to consider the application of knowledge points in actual teaching scenarios.

[0247] - Please provide structured analysis results for each knowledge point to be analyzed:

[0248] 1. Name of knowledge point to be analyzed

[0249] 2. Basis for judgment: Explain the reasons for the judgment in detail.

[0250] 3. Judgment result: Yes / No

[0251] The iterative pre-analysis agent also analyzes knowledge points by calling a large language model. Specifically, the iterative pre-analysis agent stores a preset fourth prompt word template. Based on this fourth prompt word template, the agent constructs a fourth prompt word using knowledge points selected from the set of similar knowledge points, the knowledge points that are identified as the predecessors of the target knowledge point, and the target knowledge point. The fourth prompt word is then input into the large language model to obtain the analysis results output by the model. These results indicate whether the knowledge points selected from the set of similar knowledge points and the knowledge points that are identified as the predecessors of the target knowledge point are indeed the predecessors of the target knowledge point.

[0252] For example, the requirement description in the fourth prompt word template can be as follows:

[0253] You are an experienced teaching and research teacher. Your task is to determine whether unassessed knowledge points and assessed knowledge points are prerequisite knowledge points for the target knowledge points.

[0254] Analysis steps:

[0255] 1. Necessary dependency:

[0256] - Does the definition or application of the target knowledge point directly use the prerequisite knowledge point?

[0257] - If this prerequisite knowledge point is removed, can the target knowledge point still be fully understood and applied?

[0258] Note: The order in which knowledge is learned does not necessarily establish a prerequisite knowledge relationship.

[0259] 2. Concept Construction:

[0260] - Are the core concepts of the target knowledge point directly built upon the prerequisite knowledge points?

[0261] - Do the prerequisite knowledge points include the basic concepts necessary for understanding the target knowledge point?

[0262] This conceptual dependency must be clearly reflected in the definition of the target knowledge point.

[0263] 3. Practical Applications:

[0264] - When applying the target knowledge points, is it necessary to use the prerequisite knowledge points?

[0265] These methods or skills must be an indispensable part of applying the target knowledge points.

[0266] - Rather than a dispensable auxiliary tool

[0267] 4. Direct association:

[0268] - Are there direct conceptual references or methodological applications between the prerequisite knowledge points and the target knowledge points?

[0269] - Avoid misclassifying indirectly related knowledge points or knowledge points that only appear earlier in the teaching sequence as prerequisite knowledge points.

[0270] - Focus on the substantive dependencies between concepts

[0271] 5. Review of identified knowledge points

[0272] - Please check whether the identified prerequisite knowledge points are indeed prerequisite knowledge points for the target knowledge point.

[0273] - If you find any inappropriate information that needs to be deleted, please explain why.

[0274] Precautions:

[0275] - The judgment is based strictly on the definition of the target knowledge points, rather than on the traditional teaching sequence.

[0276] - Prerequisite knowledge points must substantially help in mastering the target knowledge point; without mastering the prerequisite knowledge points, it is impossible to master the target knowledge point.

[0277] - Simply having a specific learning order is not a sufficient condition for recognizing prerequisite knowledge points.

[0278] - Remain objective and explain your analysis in detail.

[0279] - The analysis needs to consider the application of knowledge points in actual teaching scenarios.

[0280] - Please provide structured analysis results for the identified knowledge points:

[0281] 1. The name of the knowledge point has been identified.

[0282] 2. Judgment Basis: Output the reasons for your judgment after independent reflection.

[0283] 3. Judgment result: Keep / Delete

[0284] - Please provide structured analysis results for each unjudged knowledge point:

[0285] 1. The name of the knowledge point was not determined.

[0286] 2. Basis for judgment: Explain the reasons for the judgment in detail.

[0287] 3. Judgment result: Yes / No

[0288] After obtaining the prerequisite knowledge points determined by the pre-analysis agent, the pre-verification agent can be invoked to verify these prerequisite knowledge points. The pre-verification agent stores a preset fifth prompt word template. Based on this template, the prerequisite knowledge points determined by the pre-analysis agent, and the target knowledge point, the pre-verification agent constructs a fifth prompt word. This fifth prompt word is then input into the large language model, and the analysis result output by the large language model is obtained. This analysis result indicates whether the input prerequisite knowledge point is indeed a true prerequisite knowledge point for the target knowledge point.

[0289] For example, the requirement description in the fifth prompt word template can be as follows:

[0290] You are a seasoned curriculum research expert, and you are now required to verify whether the current prerequisite knowledge points are indeed prerequisite knowledge points for the target knowledge points. Please note that your task is to conduct a more in-depth verification and review of the preliminary analysis results.

[0291] Please analyze the overall picture from the following dimensions:

[0292] 1. Hierarchical relationship between knowledge points:

[0293] - Is the current prerequisite knowledge point at the level directly above the target knowledge point, rather than a cross-level relationship?

[0294] - Are there more fundamental knowledge points that serve as intermediate steps, making the current prerequisite knowledge points actually indirectly prerequisites?

[0295] 2. Verification through teaching practice:

[0296] In actual teaching, when students do not master the prerequisite knowledge points, are they truly unable to understand the target knowledge points?

[0297] - Are there any teaching cases that demonstrate that students can still understand the target knowledge points even without fully mastering the current prerequisite knowledge points?

[0298] 3. Alternative Path Analysis:

[0299] - Are there other knowledge paths that can replace the current prerequisite knowledge points for understanding the target knowledge point?

[0300] - Is the current prerequisite knowledge the only way to understand the target knowledge point?

[0301] 4. Test the independence of knowledge points:

[0302] - How much of the content within the target knowledge point can exist independently of the current prerequisite knowledge points?

[0303] - Does the core of the target knowledge point necessarily depend on the current prerequisite knowledge points?

[0304] 5. Reverse dependency check:

[0305] Does the application of the current prerequisite knowledge points also depend on certain concepts of the target knowledge points?

[0306] - Is there a circular dependency between the two rather than a one-way dependency?

[0307] 6. Final judgment result:

[0308] - Based on the analysis of all the aforementioned dimensions, a clear conclusion is drawn, with particular attention paid to any discrepancies that may exist between the conclusion and the initial analysis.

[0309] Precautions:

[0310] - The judgment is based strictly on the definition of the target knowledge point, focusing on the substantive connections between knowledge points rather than superficial similarities.

[0311] - Prerequisite knowledge points must substantially help in mastering the target knowledge point; without mastering the prerequisite knowledge points, it is impossible to master the target knowledge point.

[0312] - Simply having a specific learning order is not a sufficient condition for recognizing prerequisite knowledge points.

[0313] - Remain objective and explain your analysis in detail.

[0314] - The analysis needs to consider the application of knowledge points in actual teaching scenarios.

[0315] - Please provide structured analysis results for each prerequisite knowledge point:

[0316] 1. Names of prerequisite knowledge points

[0317] 2. Judgment basis: Specifically explain the application of the target knowledge point.

[0318] 3. Judgment result: Correct / Incorrect

[0319] It should be noted that the embodiments in this specification are merely illustrative examples of writing the target knowledge point into the prompt word every time the large language model is called. In another embodiment, considering that the large language model has memory capabilities, in the entire process of predicting the prerequisite knowledge points of the target knowledge point, the target knowledge point only needs to be written into the prompt word the first time the large language model is called. In subsequent calls to the large language model, it is not necessary to write the target knowledge point into the prompt word again.

[0320] S102. If K is greater than 1, then analyze the knowledge point chain with a prerequisite relationship from the K candidate prerequisite knowledge points. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the prerequisite relationship.

[0321] For example, if target knowledge point A has two candidate prerequisite knowledge points B and C, and if candidate prerequisite knowledge point B is a prerequisite knowledge point for candidate prerequisite knowledge point C, then learning C already includes learning B, so B needs to be removed from the candidate prerequisite knowledge points of A.

[0322] In some possible implementations, knowledge point chains with prerequisite relationships can be analyzed by generating knowledge point learning paths. This involves generating learning paths for K candidate prerequisite knowledge points to analyze whether a knowledge point chain with a prerequisite relationship exists among the K candidate prerequisite knowledge points. Analyzing the K candidate prerequisite knowledge points to identify such a chain includes: generating learning paths for the K candidate prerequisite knowledge points to obtain at least one knowledge point learning path. This at least one knowledge point learning path collectively contains all K candidate prerequisite knowledge points without repetition, and the knowledge points in each learning path are ordered according to the learning sequence. A knowledge point learning path that includes at least two candidate prerequisite knowledge points is considered a knowledge point chain with a prerequisite relationship.

[0323] For example, target knowledge point A has three candidate prerequisite knowledge points B, C, and D. Based on these three candidate prerequisite knowledge points B, C, and D, two knowledge point learning paths are generated: B→C and D. Among them, the knowledge point learning path B→C is a chain of knowledge points with prerequisite relationships. The knowledge point learning path D does not have any prerequisite relationships.

[0324] In some embodiments, learning paths for K candidate prerequisite knowledge points can be generated through a large language model, or other methods can be used to generate learning paths for K candidate prerequisite knowledge points. This specification does not limit the embodiments in this way.

[0325] S103. Based on the knowledge point chain, delete all knowledge points in the knowledge point chain except the last knowledge point in the knowledge point chain from the K candidate prerequisite knowledge points to obtain the reference prerequisite knowledge points of the target knowledge point.

[0326] In some possible implementations, a knowledge point chain with a preceding relationship can be analyzed based on a preceding merging agent. Then, based on this knowledge point chain, all knowledge points in the chain except the last one are deleted from the K candidate preceding knowledge points to obtain the reference preceding knowledge points for the target knowledge point. That is, steps S102 and S103 above include: inputting the K candidate preceding knowledge points into the preceding merging agent to analyze the knowledge point chain with a preceding relationship from the K candidate preceding knowledge points, and deleting all knowledge points in the chain except the last one from the K candidate preceding knowledge points to obtain the reference preceding knowledge points for the target knowledge point.

[0327] In some embodiments, K candidate prerequisite knowledge points are input into a prerequisite merging agent to analyze a knowledge point chain with prerequisite relationships from the K candidate prerequisite knowledge points. Based on the knowledge point chain, knowledge points other than the last knowledge point in the knowledge point chain are deleted from the K candidate prerequisite knowledge points to obtain the reference prerequisite knowledge points of the target knowledge point. This includes: inputting the K candidate prerequisite knowledge points into the prerequisite merging agent; the prerequisite merging agent constructing a sixth prompt word based on the K candidate prerequisite knowledge points; the sixth prompt word being used to instruct the large language model to analyze a knowledge point chain with prerequisite relationships from the K candidate prerequisite knowledge points; and deleting knowledge points other than the last knowledge point in the knowledge point chain from the K candidate prerequisite knowledge points; the prerequisite merging agent inputting the sixth prompt word into the large language model to obtain the analysis result output by the large language model, which includes the reference prerequisite knowledge points.

[0328] In some embodiments, the large language model can also output learning paths for each knowledge point.

[0329] In some embodiments, the pre-merging agent stores a preset sixth prompt word template. This sixth prompt word template includes: a requirement description related to analyzing a knowledge point chain with a pre-existing relationship and deleting all knowledge points in the chain except the last one; and placeholders for filling in candidate pre-existing knowledge points. The corresponding placeholders in the sixth prompt word template are replaced with candidate pre-existing knowledge points to obtain the sixth prompt word. In some embodiments, the requirement description includes a task objective description and a task execution rule description.

[0330] S104. Obtain the extended knowledge point set and check whether there is a knowledge point loop with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge point of each reference prerequisite knowledge point.

[0331] In some embodiments, a directed graph can be constructed to check whether a knowledge point cycle with a prerequisite relationship exists in the extended knowledge point set. For example, to check whether a knowledge point cycle with a prerequisite relationship exists in the extended knowledge point set, a directed graph is constructed based on the prerequisite relationships between the knowledge points in the extended knowledge point set; a query is made to check whether a bidirectional edge exists in the directed graph; and / or, a query is made to check whether a cycle exists in the directed graph.

[0332] In some embodiments, a large language model can be used to check whether a knowledge point cycle with a prerequisite relationship exists in the extended knowledge point set. It should be noted that the embodiments in this specification are merely illustrative examples of how to "check whether a knowledge point cycle with a prerequisite relationship exists in the extended knowledge point set." Of course, other methods not mentioned in this specification can also be used, and these embodiments do not impose any limitations on them.

[0333] S105. If a knowledge point cycle exists, then based on the knowledge point cycle, the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set are deleted, so as to delete the knowledge points with incorrect prerequisite relationships based on the knowledge point cycle, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point cycle.

[0334] In some possible implementations, based on the knowledge point ring, knowledge points in each reference prerequisite knowledge point and / or each reference prerequisite knowledge point's prerequisite knowledge point are reduced, including: when the knowledge point ring consists of two knowledge points, identifying the more fundamental knowledge point from the two knowledge points, and deleting the other knowledge point from the prerequisite knowledge point of the more fundamental knowledge point.

[0335] In some possible implementations, based on the knowledge point ring, knowledge points in each reference prerequisite knowledge point and / or each reference prerequisite knowledge point's prerequisite knowledge point are reduced from the knowledge points in the extended knowledge point set. This includes: when the knowledge point ring includes two or more knowledge points, analyzing the prerequisite dependencies corresponding to each prerequisite relationship in the knowledge point ring, where the prerequisite dependency degree is used to represent the necessity of mastering the prerequisite knowledge point before learning the subsequent knowledge point; identifying prerequisite relationships that do not meet the prerequisite dependency degree condition as breakpoints; deleting the prerequisite knowledge point corresponding to the breakpoint from the prerequisite knowledge points of its subsequent knowledge point from the extended knowledge point set; and / or, sorting the knowledge points in the knowledge point ring according to the learning order to obtain the target learning path; and deleting reference prerequisite knowledge points and / or the prerequisite knowledge points of each reference prerequisite knowledge point that do not conform to the target learning path from the prerequisite knowledge points of their corresponding subsequent knowledge points.

[0336] In some possible implementations, step S105 above is implemented by a loop processing agent. In some embodiments, the loop processing agent includes a simple loop processing agent and a complex loop processing agent. When the knowledge point loop consists of two knowledge points, the simple loop processing agent is invoked to identify the more fundamental knowledge point at the cognitive level from the two knowledge points, and the other knowledge point is removed from the predecessor knowledge points of the more fundamental knowledge point. When the knowledge point loop includes more than two knowledge points, the complex loop processing agent is invoked to analyze the predecessor dependencies corresponding to each predecessor relationship in the knowledge point loop, identify predecessor relationships that do not meet the predecessor dependency conditions as breakpoints, and remove the predecessor knowledge points corresponding to the breakpoints from the predecessor knowledge points of their successors from the extended knowledge point set; and / or, sort the knowledge points in the knowledge point loop according to the learning order to obtain the target learning path, and remove the reference predecessor knowledge points and / or the predecessor knowledge points of each reference predecessor knowledge point that do not conform to the target learning path from the predecessor knowledge points of their corresponding successors.

[0337] In some embodiments, a simple loop processing agent is invoked to identify a more fundamental knowledge point from two knowledge points, and to remove the other knowledge point from the preceding knowledge points of the more fundamental knowledge point. This includes: inputting two knowledge points forming a knowledge point loop into the simple loop processing agent; the simple loop processing agent constructing a seventh prompt word based on these two knowledge points, which instructs the large language model to identify the more fundamental knowledge point from these two knowledge points; the simple loop processing agent inputting the seventh prompt word into the large language model to obtain the analysis result output by the large language model, which includes the more fundamental knowledge point; and the simple loop processing agent removing the other knowledge point from the preceding knowledge points of the more fundamental knowledge point.

[0338] In some embodiments, the simple loop processing agent stores a preset seventh prompt word template, which includes: a requirement description related to identifying more basic knowledge points at the cognitive level, and placeholders for filling in knowledge point rings; the seventh prompt word is obtained by replacing the corresponding placeholders in the seventh prompt word template with knowledge point rings. In some embodiments, the requirement description includes a task purpose description and a task execution rule description.

[0339] In some embodiments, the process of calling a complex loop processing agent to process a knowledge point loop includes: inputting the knowledge point loop into the complex loop processing agent; the complex loop processing agent constructing an eighth prompt word based on the knowledge point loop; the eighth prompt word being used to instruct the large language model to analyze the pre-dependent procedures corresponding to each pre-dependent relation in the knowledge point loop; identifying pre-dependent relations that do not meet the pre-dependent degree condition as breakpoints; deleting the pre-dependent knowledge points corresponding to the breakpoints from the pre-dependent knowledge points of their successors from the extended knowledge point set; sorting the knowledge points in the knowledge point loop according to the learning order to obtain the target learning path; the complex loop processing agent inputting the eighth prompt word into the large language model to obtain the analysis results output by the large language model, which include the target learning path; and the complex loop processing agent deleting reference pre-dependent knowledge points and / or the pre-dependent knowledge points of each reference pre-dependent knowledge point that do not conform to the target learning path from the pre-dependent knowledge points of their corresponding successors.

[0340] In some embodiments, the complex loop processing agent stores a preset eighth prompt word template, which includes: a requirement description related to analyzing the breakpoint of the knowledge point loop and sorting the knowledge points in the knowledge point loop according to the learning order; and placeholders for filling in the knowledge point loops respectively. The eighth prompt word is obtained by replacing the corresponding placeholders in the eighth prompt word template with the knowledge point loops. In some embodiments, the requirement description includes a task purpose description and a task execution rule description.

[0341] For example, such as Figure 5 As shown, the candidate prior knowledge points output by the prior analysis agent are input into the prior merging agent. The prior merging agent constructs the sixth prompt word based on the candidate prior knowledge points and the sixth prompt word template, and inputs the sixth prompt word into the large language model to obtain the analysis results output by the large language model.

[0342] The requirement description in the sixth prompt word template can be as follows:

[0343] You are a senior teaching and research expert, and you need help to consolidate the sequential relationships of a group of knowledge points.

[0344] Task Description:

[0345] Given a set of knowledge points, you need to analyze the relationships between these knowledge points, find the highest-level knowledge point, and remove its preceding knowledge points.

[0346] Merge method:

[0347] 1. Analyze the pre- and post-dependencies between given knowledge points.

[0348] - Suppose there are three knowledge points A, B, and C, and there exists a sequential relationship of A -> B -> C.

[0349] 2. Deletion Rules

[0350] - A and B are prerequisite knowledge points for C.

[0351] - Then delete A and B, keeping only C.

[0352] Reason: C is the highest level of knowledge; mastering C means you have already mastered A and B.

[0353] Precautions:

[0354] - The judgment is based strictly on the definition of knowledge points, rather than on the traditional teaching sequence.

[0355] - Prerequisite knowledge points must be directly reflected in the definition or application of subsequent knowledge points.

[0356] - Simply having a specific learning order is not a sufficient condition for recognizing prerequisite knowledge points.

[0357] - Remain objective and explain your analysis in detail.

[0358] - The analysis needs to consider the application of knowledge points in actual teaching scenarios.

[0359] - The merging conclusion needs to be provided as follows:

[0360] 1. Retained knowledge points

[0361] 2. Precedence / Precedence Relationship: The preceding and following relationships of knowledge points (connected by ->).

[0362] 3. Deleted knowledge points: Knowledge points preceding the retained knowledge points in the sequence (fill in "None" if none).

[0363] Next, the loop processing agent obtains an extended knowledge point set based on the output of the preceding analysis agent. It checks whether there is a knowledge point loop with a preceding relationship in the extended knowledge point set. If a knowledge point loop exists, and the knowledge point loop consists of two knowledge points, the simple loop processing agent is called. The simple loop processing agent constructs a seventh prompt word based on the knowledge point loop and the seventh prompt word template, and inputs the seventh prompt word into the large language model to obtain the analysis result output by the large language model. This analysis result is used to represent the preceding and following relationship between the two knowledge points.

[0364] The requirement description in the seventh prompt word template can be as follows:

[0365] You are a senior teaching and research expert who needs help to solve the circular dependency problem between two knowledge points and determine which knowledge point is a prerequisite and which is a subsequent knowledge point.

[0366] Prerequisite knowledge points for judgment:

[0367] 1. Fundamental Concepts:

[0368] - The prerequisite knowledge points should include more basic and general concepts.

[0369] - The core concepts of subsequent knowledge points should be built upon the foundation of prior knowledge points.

[0370] - Prerequisite knowledge should be a necessary prerequisite for understanding subsequent knowledge.

[0371] 2. Logical dependency:

[0372] - The definition or derivation of subsequent knowledge points must directly depend on the preceding knowledge points.

[0373] - If prerequisite knowledge is lacking, the understanding of subsequent knowledge will show logical gaps.

[0374] This dependency should be irreversible.

[0375] 3. Necessity of application:

[0376] - When applying subsequent knowledge points in practice, it is necessary to use the preceding knowledge points.

[0377] - The methods or principles of prerequisite knowledge are necessary tools when applying subsequent knowledge.

[0378] This application relationship should be direct, not indirect.

[0379] 4. Cognitive hierarchy:

[0380] - The prerequisite knowledge points should be at a more basic cognitive level.

[0381] - Subsequent knowledge points should be extensions or deepenings based on prior knowledge points.

[0382] - This hierarchical relationship should conform to the laws of cognitive development.

[0383] 5. Practicality in teaching:

[0384] In actual teaching scenarios, prerequisite knowledge points should be taught before subsequent knowledge points.

[0385] This teaching sequence should conform to the cognitive patterns of most learners.

[0386] Successful cases in teaching practice should support this sequential relationship.

[0387] When the knowledge point loop includes more than two knowledge points, the complex loop processing agent is invoked. The complex loop processing agent constructs the eighth prompt word based on the knowledge point loop and the eighth prompt word template, inputs the eighth prompt word into the large language model, and obtains the analysis results output by the large language model. These analysis results are used to represent the target learning path.

[0388] The requirement description in the eighth prompt word template can be as follows:

[0389] Task: Refactor the learning path for knowledge points with circular dependencies.

[0390] Background: When designing learning paths, we encountered the problem of circular dependencies between knowledge points (such as the form A→B→C→D→A). This circular dependency can lead to a disordered learning order and affect learning effectiveness.

[0391] Objective: To create an optimized, non-cyclical learning path that ensures the logicality and effectiveness of the learning process.

[0392] enter:

[0393] 1. A series of interconnected knowledge points, each of which includes a name and definition.

[0394] 2. The existing dependencies between these knowledge points include circular dependencies.

[0395] Require:

[0396] 1. Analyze the given knowledge points and their dependencies, and clearly identify all circular dependency chains.

[0397] 2. For each circular dependency, determine the most reasonable breakpoint and explain the reason for choosing that breakpoint.

[0398] 3. Reorder the knowledge points to create a logically clear learning path.

[0399] 4. Ensure that the new learning path covers all original knowledge points.

[0400] 5. Provide a restructured learning path, including an explanation of each step.

[0401] Circular dependency handling strategies:

[0402] 1. Identify which dependencies in the loop are "strong dependencies" (must be learned first) and which are "weak dependencies" (helpful but not required).

[0403] 2. Consider breaking down certain knowledge points into basic and advanced parts to break the cycle.

[0404] 3. Evaluate which knowledge point is most suitable as the entry point for the loop (with the fewest prerequisites).

[0405] Output format:

[0406]

Circular Dependency Analysis

[0407] - All identified circular dependency chains

[0408] - Suggested breakpoints and reasons for each loop

[0409]

analyze

[0410] 1. The restructured learning path list is arranged in the order of learning.

[0411] 2. A brief explanation of each knowledge point, clarifying its position and importance in the new path.

[0412] 3. Provide reasons for any significant changes.

[0413] S106. Based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion, determine the target prerequisite knowledge points for the target knowledge point.

[0414] In some possible implementations, the reference prerequisite knowledge points in the expanded knowledge point set after the deletion operation can be directly used as the target prerequisite knowledge points of the target knowledge point. Specifically, determining the target prerequisite knowledge points of the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after the deletion operation includes: determining the reference prerequisite knowledge points in the expanded knowledge point set after the deletion operation as the target prerequisite knowledge points of the target knowledge point.

[0415] In some possible implementations, further verification by a pre-evaluation agent is required to determine the target prerequisite knowledge point. Specifically, determining the target prerequisite knowledge point based on reference prerequisite knowledge points in the expanded knowledge point set after deletion includes: inputting reference prerequisite knowledge points from the expanded knowledge point set after deletion into the pre-evaluation agent to determine whether the input knowledge points are prerequisite knowledge points for the target knowledge point; wherein the model used by the pre-evaluation agent is different from the model used to select candidate prerequisite knowledge points; deleting reference prerequisite knowledge points determined by the pre-evaluation agent to be non-prerequisite knowledge points, and retaining the reference prerequisite knowledge points determined to be prerequisite knowledge points as the target prerequisite knowledge points for the target knowledge point.

[0416] In some embodiments, the model used by the pre-evaluation agent has more complex parameters than the model used to select candidate pre-knowledge points, and therefore, the accuracy of the pre-evaluation agent is higher.

[0417] In some embodiments, the pre-evaluation agent can not only identify more accurate pre-knowledge points, but also evaluate the processing performance of other agents. For example, after determining whether the reference pre-knowledge points are correct, the pre-evaluation agent can divide the number of correct reference pre-knowledge points by the total number of all reference pre-knowledge points to obtain the pre-knowledge point accuracy. It can also construct a knowledge point learning path with the target knowledge point as its endpoint based on the reference pre-knowledge points in the expanded knowledge point set after deletion operations. If any knowledge point learning path includes an incorrect reference pre-knowledge point that the pre-evaluation agent determines is not a pre-knowledge point, then the knowledge point learning path is determined to be incorrect; otherwise, the knowledge point learning path is correct. The number of correct knowledge point learning paths can be divided by the total number of knowledge point learning paths to obtain the knowledge point learning path accuracy. Subsequently, based on the pre-knowledge point accuracy and the knowledge point learning path accuracy, it can be determined whether to optimize other agents.

[0418] In some embodiments, after determining the target prerequisite knowledge points of the target knowledge point, the target prerequisite knowledge points can be imported into the corresponding knowledge point graph database for use by downstream businesses.

[0419] In this embodiment, based on the target knowledge point itself, its initial set of prerequisite knowledge points, and the set of basic knowledge points, similar knowledge points are recalled to obtain a set of similar knowledge points, from which prerequisite knowledge points are selected. This multi-source recall mechanism significantly expands the search range of prerequisite knowledge points, thus laying a solid foundation for accurately obtaining candidate prerequisite knowledge points. Based on this, for the initially selected candidate prerequisite knowledge points, the chain of knowledge points with prerequisite relationships is further analyzed. By deleting knowledge points in this chain except for the last one, indirect prerequisite knowledge points are removed, while direct prerequisite knowledge points are retained, further improving the accuracy of prerequisite knowledge points. Subsequently, by constructing an expanded set of knowledge points and detecting possible knowledge point loops with prerequisite relationships, erroneous knowledge points that generate loops are dynamically deleted, achieving self-consistency verification and correction of the prerequisite relationship topology. The above multi-stage collaborative optimization mechanism not only enhances the comprehensiveness of knowledge point recall but also improves the accuracy of prerequisite knowledge point prediction layer by layer through a dual strategy of chain simplification and loop elimination.

[0420] Corresponding to the embodiments of the foregoing methods, this specification also provides embodiments of the apparatus and the terminal to which it is applied.

[0421] The embodiments of the document processing apparatus described in this specification can be applied to computer devices, such as servers or terminal devices. The apparatus embodiments can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logically defined apparatus, it is formed by the processor in which it processes the file, reading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 6 The diagram shown is a hardware structure diagram of a computer device containing the file processing apparatus as described in this specification, except... Figure 6 In addition to the processor 610, memory 630, network interface 620, and non-volatile memory 640 shown, the server or electronic device where the device 631 is located in the embodiment may also include other hardware depending on the actual function of the computer device, which will not be described in detail here.

[0422] like Figure 7 As shown, Figure 7 This is a block diagram illustrating a prior knowledge point prediction device according to an exemplary embodiment of this specification, the device comprising:

[0423] The candidate selection unit 701 is used to select the prerequisite knowledge points of the target knowledge point from the set of similar knowledge points to obtain K candidate prerequisite knowledge points, where K is any positive integer; the set of similar knowledge points includes knowledge points in the knowledge point library that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the set of basic knowledge points; wherein, the initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points of the target knowledge point; and the set of basic knowledge points includes the predicted basic knowledge points that need to be mastered to master the target knowledge point.

[0424] The direct pre-filtering unit 702 is used to analyze the knowledge point chains with pre-relationships from the K candidate pre-relationship knowledge points if K is greater than 1. The knowledge point chains contain at least two knowledge points and the knowledge points in the knowledge point chains are arranged in the order of the pre-relationships. Based on the knowledge point chains, the knowledge points in the knowledge point chains except for the last knowledge point in the knowledge point chains are deleted from the K candidate pre-relationship knowledge points to obtain the reference pre-relationship knowledge points of the target knowledge point.

[0425] The cyclic decomposition unit 703 is used to obtain an extended knowledge point set and check whether there is a knowledge point loop with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point loop exists, the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set are deleted based on the knowledge point loop, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point loop.

[0426] The target determination unit 704 is used to determine the target prerequisite knowledge points of the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion.

[0427] In some possible implementations, the candidate selection unit 701 is specifically used for:

[0428] Select N knowledge points from the set of similar knowledge points, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are the pre-knowledge points of the target knowledge points;

[0429] When the L input knowledge points are determined to be prerequisite knowledge points for the target knowledge point, if there are still unselected knowledge points in the set of similar knowledge points, then M knowledge points are selected from the unselected knowledge points in the set of similar knowledge points, and these M knowledge points and the L knowledge points are taken as the current knowledge points. The process returns to the step of inputting the current knowledge points to the pre-analysis agent to determine whether the M knowledge points are prerequisite knowledge points for the target knowledge point. Wherein, when at least one of the M knowledge points is more suitable as a prerequisite knowledge point for the target knowledge point than at least one of the L knowledge points, at least one of the L knowledge points is determined not to be a prerequisite knowledge point for the target knowledge point, where M is an integer less than or equal to N, and L is an integer greater than or equal to 1.

[0430] Based on the prior knowledge points determined by the prior analysis agent, the K candidate prior knowledge points are obtained.

[0431] In some possible implementations, the candidate selection unit 701 is used to determine the prerequisite knowledge points identified by the prerequisite analysis agent as candidate prerequisite knowledge points, thereby obtaining the K candidate prerequisite knowledge points; or,

[0432] The candidate selection unit 701 is used to input the prerequisite knowledge points determined by the prerequisite analysis agent into the prerequisite verification agent to verify whether the input knowledge points are prerequisite knowledge points of the target knowledge points. The prerequisite verification agent and the prerequisite analysis agent have different judgment dimensions for the prerequisite knowledge points. The prerequisite knowledge points that pass the verification are used as candidate prerequisite knowledge points to obtain the K candidate prerequisite knowledge points.

[0433] In some possible implementations, the direct pre-screening unit 702 is used to generate learning paths for the K candidate pre-knowledge points to obtain at least one knowledge point learning path, wherein the at least one knowledge point learning path collectively contains the K candidate pre-knowledge points without repetition, and the knowledge points in each knowledge point learning path are arranged in the order of learning.

[0434] Among them, the knowledge point learning path that includes at least two candidate prerequisite knowledge points is the knowledge point chain with a prerequisite relationship.

[0435] In some possible implementations, the cyclic disassembly unit 703 is specifically used for:

[0436] When the knowledge point loop consists of two knowledge points, the knowledge point with a more basic cognitive level is identified from the two knowledge points, and the other knowledge point is deleted from the preceding knowledge points of the more basic cognitive level knowledge point.

[0437] And / or,

[0438] When the knowledge point loop includes two or more knowledge points, the degree of prerequisite dependency corresponding to each prerequisite relationship in the knowledge point loop is analyzed. The degree of prerequisite dependency is used to represent the necessity of mastering the prerequisite knowledge points before learning the subsequent knowledge points. Prerequisite relationships that do not meet the prerequisite dependency condition are identified as breakpoints. From the extended knowledge point set, the prerequisite knowledge points corresponding to the breakpoints are deleted from the prerequisite knowledge points of their subsequent knowledge points. And / or, the knowledge points in the knowledge point loop are sorted according to the learning order to obtain the target learning path. Reference prerequisite knowledge points and / or the prerequisite knowledge points of each reference prerequisite knowledge point that do not conform to the target learning path are deleted from the prerequisite knowledge points of the corresponding knowledge points.

[0439] In some possible implementations, the target determination unit 704 is specifically used to input reference prerequisite knowledge points from the expanded knowledge point set after deletion operations into a prerequisite evaluation agent to determine whether the input knowledge points are prerequisite knowledge points of the target knowledge point; wherein, the model used by the prerequisite evaluation agent is different from the model used to select candidate prerequisite knowledge points; reference prerequisite knowledge points that are determined by the prerequisite evaluation agent to be non-prerequisite knowledge points are deleted, and reference prerequisite knowledge points that are determined to be prerequisite knowledge points are retained as target prerequisite knowledge points of the target knowledge point.

[0440] like Figure 8 As shown, in some possible implementations, the device further includes:

[0441] The knowledge point recall unit 705 is used to query knowledge points in the knowledge point database that meet the similarity conditions to the target knowledge point, and determine the score of the queried knowledge point based on the similarity corresponding to the queried knowledge point.

[0442] The knowledge point recall unit 705 is further configured to query knowledge points in the knowledge point database that meet the similarity conditions for each knowledge point in the initial set of prerequisite knowledge points, and to determine the score of the queried knowledge point by multiplying the similarity corresponding to the queried knowledge point with the first weight coefficient.

[0443] The knowledge point recall unit 705 is further configured to query knowledge points in the knowledge point database that meet the similarity conditions for each knowledge point in the basic knowledge point set, and determine the similarity and second weight coefficient of the queried knowledge points as the score of each queried knowledge point, wherein the first weight coefficient and the second weight coefficient are both greater than 1, and the first weight coefficient is greater than the second weight coefficient.

[0444] The knowledge point recall unit 705 is also used to identify knowledge points whose scores exceed a preset threshold among the various knowledge points queried as knowledge points in the set of similar knowledge points.

[0445] In some possible implementations, the knowledge point recall unit 705 is specifically used for:

[0446] Based on the name of the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the name of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0447] Based on the definition of the target knowledge point, query the knowledge point database for knowledge points whose definitions meet the similarity conditions with the definition of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points;

[0448] Based on the concepts involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the concepts involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0449] Based on the professional terms involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the professional terms involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient.

[0450] The concepts and professional terms involved in the target knowledge points are obtained through semantic parsing based on the names and definitions of the target knowledge points; each concept and professional term has a corresponding importance parameter, which is used to represent the importance of the corresponding concept or professional term in the target knowledge point; the third weight coefficient is determined based on the importance parameter, and the third weight coefficient is positively correlated with the importance parameter and is greater than 1.

[0451] Accordingly, this specification also provides an apparatus comprising a processor; a memory for storing processor-executable instructions; wherein the processor is configured to:

[0452] From the set of similar knowledge points, select the prerequisite knowledge points for the target knowledge point to obtain K candidate prerequisite knowledge points, where K is any positive integer; the set of similar knowledge points includes knowledge points in the knowledge point library that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the basic knowledge point set; wherein, the initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points for the target knowledge point; the basic knowledge point set includes the predicted basic knowledge points that need to be mastered to master the target knowledge point;

[0453] If K is greater than 1, then analyze the knowledge point chains with prerequisite relationships from the K candidate prerequisite knowledge points. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the prerequisite relationship. Based on the knowledge point chain, delete the knowledge points in the knowledge point chain except for the last knowledge point in the knowledge point chain from the K candidate prerequisite knowledge points to obtain the reference prerequisite knowledge points of the target knowledge point.

[0454] Obtain an extended knowledge point set, and check whether there is a knowledge point cycle with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point cycle exists, then based on the knowledge point cycle, delete the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set, so as to delete the knowledge points with incorrect prerequisite relationships based on the knowledge point cycle, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point cycle.

[0455] Based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion, the target prerequisite knowledge points of the target knowledge point are determined.

[0456] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0457] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the solution in this specification according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0458] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0459] Other embodiments of this specification will readily occur to those skilled in the art upon consideration of the specification and practice of the invention claimed herein. This specification is intended to cover any variations, uses, or adaptations that follow the general principles of this specification and include common knowledge or customary techniques in the art not claimed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this specification are indicated by the following claims.

[0460] It should be understood that this specification is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this specification is limited only by the appended claims.

[0461] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for predicting prerequisite knowledge points, characterized in that, The method includes: The agent performs a pre-analysis to determine the prerequisite knowledge points for the target knowledge point from a set of similar knowledge points, resulting in K candidate prerequisite knowledge points, where K is any positive integer. The set of similar knowledge points includes knowledge points in the knowledge point database that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the basic knowledge point set. The initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points for the target knowledge point. The basic knowledge point set includes the predicted basic knowledge points that need to be mastered to master the target knowledge point. If K is greater than 1, then analyze the knowledge point chains with prerequisite relationships from the K candidate prerequisite knowledge points. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the prerequisite relationship. Based on the knowledge point chain, delete the knowledge points in the knowledge point chain except for the last knowledge point in the knowledge point chain from the K candidate prerequisite knowledge points to obtain the reference prerequisite knowledge points of the target knowledge point. Obtain an extended knowledge point set, and check whether there is a knowledge point cycle with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point cycle exists, then based on the knowledge point cycle, delete the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set, so as to delete the knowledge points with incorrect prerequisite relationships based on the knowledge point cycle, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point cycle. Based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion, the target prerequisite knowledge points of the target knowledge point are determined.

2. The method according to claim 1, characterized in that, The step of determining the prerequisite knowledge points for the target knowledge point from the set of similar knowledge points through the pre-analysis agent includes: Select N knowledge points from the set of similar knowledge points, take these N knowledge points as the current knowledge points, and input the current knowledge points into the pre-analysis agent to determine whether the input knowledge points are the pre-knowledge points of the target knowledge points; When the L input knowledge points are determined to be prerequisite knowledge points for the target knowledge point, if there are still unselected knowledge points in the set of similar knowledge points, then M knowledge points are selected from the unselected knowledge points in the set of similar knowledge points, and these M knowledge points and the L knowledge points are taken as the current knowledge points. The process returns to the step of inputting the current knowledge points to the pre-analysis agent to determine whether the M knowledge points are prerequisite knowledge points for the target knowledge point. Wherein, when at least one of the M knowledge points is more suitable as a prerequisite knowledge point for the target knowledge point than at least one of the L knowledge points, at least one of the L knowledge points is determined not to be a prerequisite knowledge point for the target knowledge point, where M is an integer less than or equal to N, and L is an integer greater than or equal to 1. Based on the prior knowledge points determined by the prior analysis agent, the K candidate prior knowledge points are obtained.

3. The method according to claim 2, characterized in that, The process of obtaining the K candidate pre-knowledge points based on the pre-analysis agent includes: The prerequisite knowledge points determined by the preceding analysis agent are selected as candidate prerequisite knowledge points, resulting in the K candidate prerequisite knowledge points; or... The prerequisite knowledge points determined by the prerequisite analysis agent are input into the prerequisite verification agent to verify whether the input knowledge points are prerequisite knowledge points of the target knowledge points. The prerequisite verification agent and the prerequisite analysis agent have different judgment dimensions for prerequisite knowledge points. The prerequisite knowledge points that pass the verification are used as candidate prerequisite knowledge points to obtain the K candidate prerequisite knowledge points.

4. The method according to claim 1, characterized in that, The step of analyzing the chain of knowledge points with preceding relationships from the K candidate prior knowledge points includes: Learning paths are generated for the K candidate prerequisite knowledge points to obtain at least one knowledge point learning path. The at least one knowledge point learning path contains all K candidate prerequisite knowledge points without repetition, and the knowledge points in each knowledge point learning path are arranged in the order of learning. Among them, the knowledge point learning path that includes at least two candidate prerequisite knowledge points is the knowledge point chain with a prerequisite relationship.

5. The method according to claim 1, characterized in that, The step of reducing the number of knowledge points in each reference prerequisite knowledge point and / or each reference prerequisite knowledge point's prerequisite knowledge point from the extended knowledge point set based on the knowledge point ring includes: When the knowledge point loop consists of two knowledge points, the knowledge point with a more basic cognitive level is identified from the two knowledge points, and the other knowledge point is deleted from the preceding knowledge points of the more basic cognitive level knowledge point. When the knowledge point loop includes two or more knowledge points, the degree of prerequisite dependency corresponding to each prerequisite relationship in the knowledge point loop is analyzed. The degree of prerequisite dependency is used to represent the necessity of mastering the prerequisite knowledge points before learning the subsequent knowledge points. Prerequisite relationships that do not meet the prerequisite dependency condition are identified as breakpoints. From the extended knowledge point set, the prerequisite knowledge points corresponding to the breakpoints are deleted from the prerequisite knowledge points of their subsequent knowledge points. And / or, the knowledge points in the knowledge point loop are sorted according to the learning order to obtain the target learning path. Reference prerequisite knowledge points and / or the prerequisite knowledge points of each reference prerequisite knowledge point that do not conform to the target learning path are deleted from the prerequisite knowledge points of the corresponding knowledge points.

6. The method according to claim 1, characterized in that, The process of determining the target prerequisite knowledge points for the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion includes: The reference prerequisite knowledge points in the expanded knowledge point set after the deletion operation are input into the prerequisite evaluation agent to determine whether the input knowledge points are prerequisite knowledge points of the target knowledge point; wherein, the model used by the prerequisite evaluation agent is different from the model used to select candidate prerequisite knowledge points; Delete the reference prior knowledge points that are not prior knowledge points as determined by the prior evaluation agent, and retain the reference prior knowledge points that are determined to be prior knowledge points as the target prior knowledge points of the target knowledge point.

7. The method according to claim 1, characterized in that, The process of generating the set of similar knowledge points includes: For the target knowledge point, query the knowledge point database for knowledge points that meet the similarity conditions to the target knowledge point, and determine the score of the queried knowledge point based on the similarity corresponding to the queried knowledge point; For each knowledge point in the initial set of prerequisite knowledge points, query the knowledge point database for knowledge points that meet the similarity conditions with the knowledge point, and multiply the similarity of the queried knowledge point by the first weight coefficient to determine the score of the queried knowledge point. For each knowledge point in the set of basic knowledge points, query the knowledge point database for knowledge points that meet the similarity conditions with the knowledge point, and determine the similarity and second weight coefficient of the queried knowledge points as the score of each queried knowledge point. The first weight coefficient and the second weight coefficient are both greater than 1, and the first weight coefficient is greater than the second weight coefficient. Knowledge points whose scores exceed a preset threshold among the retrieved knowledge points are identified as knowledge points in the set of similar knowledge points.

8. The method according to claim 7, characterized in that, The step of querying knowledge points from the knowledge point database that meet similarity criteria to the target knowledge point, and determining the score of the queried knowledge points based on their similarity, includes: Based on the name of the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the name of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points; Based on the definition of the target knowledge point, query the knowledge point database for knowledge points whose definitions meet the similarity conditions with the definition of the target knowledge point, and determine the similarity of the queried knowledge points as the score of the queried knowledge points; Based on the concepts involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the concepts involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient. Based on the professional terms involved in the target knowledge point, query the knowledge point database for knowledge points whose names meet the similarity conditions with the professional terms involved in the target knowledge point, and determine the score of the query knowledge point by multiplying the similarity corresponding to the query knowledge point with the third weight coefficient. The concepts and professional terms involved in the target knowledge points are obtained through semantic parsing based on the names and definitions of the target knowledge points; each concept and professional term has a corresponding importance parameter, which is used to represent the importance of the corresponding concept or professional term in the target knowledge point; the third weight coefficient is determined based on the importance parameter, and the third weight coefficient is positively correlated with the importance parameter and is greater than 1.

9. A pre-knowledge point prediction device, characterized in that, The device includes: The candidate selection unit is used to determine the prerequisite knowledge points of the target knowledge point from the set of similar knowledge points through the pre-analysis agent, obtaining K candidate prerequisite knowledge points, where K is any positive integer; the set of similar knowledge points includes knowledge points in the knowledge point database that are similar to the target knowledge point, each knowledge point in the initial set of prerequisite knowledge points, and each knowledge point in the basic knowledge point set; wherein, the initial set of prerequisite knowledge points includes the predicted prerequisite knowledge points of the target knowledge point; the basic knowledge point set includes the predicted basic knowledge points that need to be mastered to master the target knowledge point; A direct pre-filtering unit is used to analyze knowledge point chains with pre-relationships from the K candidate pre-relationship knowledge points if K is greater than 1. The knowledge point chain contains at least two knowledge points and the knowledge points in the knowledge point chain are arranged in the order of the pre-relationship. Based on the knowledge point chain, knowledge points other than the last knowledge point in the knowledge point chain are deleted from the K candidate pre-relationship knowledge points to obtain the reference pre-relationship knowledge points of the target knowledge point. The cyclic decomposition unit is used to obtain an extended knowledge point set and check whether there is a knowledge point loop with a prerequisite relationship in the extended knowledge point set. The extended knowledge point set includes the target knowledge point, each obtained reference prerequisite knowledge point, and the prerequisite knowledge points of each reference prerequisite knowledge point. If the knowledge point loop exists, the knowledge points in each reference prerequisite knowledge point and / or the prerequisite knowledge points of each reference prerequisite knowledge point in the extended knowledge point set are deleted based on the knowledge point loop, so that the knowledge points in the extended knowledge point set after the deletion operation do not have a knowledge point loop. The target determination unit is used to determine the target prerequisite knowledge points of the target knowledge point based on the reference prerequisite knowledge points in the expanded knowledge point set after deletion.

10. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.

12. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the method described in any one of claims 1 to 8.

Images