CN118657140A - A test question knowledge point annotation method and system based on artificial intelligence - Google Patents

Abstract

The invention discloses a test question knowledge point labeling method and a system based on artificial intelligence, which belong to the technical field of test question knowledge point labeling, wherein the method comprises the following steps: collecting test question data to be marked, wherein the test question data comprises a question text and a plurality of answer texts; sequentially carrying out knowledge point contribution analysis processing on the multiple answer texts to obtain multiple knowledge point contribution scores; and integrating a plurality of knowledge point contribution scores corresponding to each answer text based on the topic text to obtain multi-solution knowledge point labeling data. According to the invention, the comprehensiveness of knowledge point labeling is obviously improved by introducing the knowledge point labeling strategy of the multiple solutions, and by analyzing the multiple solution strategies, not only can the unique knowledge points behind different solutions be captured, but also the internal relation between the knowledge points can be revealed, so that a multi-angle learning view angle is provided for students, and the cultivation of criticizing thinking and innovation capability is promoted.

Description

A test question knowledge point annotation method and system based on artificial intelligence

Technical Field

The present invention relates to the technical field of exercise annotation, and in particular to an artificial intelligence-based test question knowledge point annotation method and system.

Background Art

Currently, in the field of educational technology, the annotation of knowledge points in test questions mainly focuses on the analysis of a single solution. Although this approach can provide more in-depth knowledge point annotation for a specific problem-solving path, it ignores the diversity and complexity of problem solving. In real educational scenarios, many problems often have multiple solutions, and each solution contains unique knowledge points and ways of thinking. However, the traditional single solution annotation method often only focuses on the most common or most direct way to solve the problem, which leads to the neglect of other potential knowledge points and limits the expansion of students' thinking and the cultivation of creativity.

Summary of the invention

In order to solve at least one of the technical problems raised above, the present invention provides a test question knowledge point labeling method based on artificial intelligence, which significantly improves the comprehensiveness of knowledge point labeling by introducing a knowledge point labeling strategy with multiple solutions.

In one aspect, the present invention provides a method for labeling test knowledge points based on artificial intelligence, the method comprising:

Collecting test data to be annotated, the test data including a question text and a plurality of answer texts, each answer text corresponding to a different problem-solving strategy;

Performing knowledge point contribution analysis on the plurality of answer texts in sequence to obtain a plurality of knowledge point contribution scores;

Based on the question text, the contribution scores of the multiple knowledge points corresponding to each of the answer texts are integrated to obtain multi-solution knowledge point annotation data.

Preferably, the step of sequentially performing knowledge point contribution analysis on the plurality of answer texts to obtain a plurality of knowledge point contribution scores includes:

Performing knowledge point type word segmentation processing on the answer text to obtain multiple knowledge point type text information sets and multiple argumentation connectives;

Obtaining knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and matching corresponding contribution weights from a preset argumentation connective contribution weight table based on a plurality of the argument connectives to obtain logical contribution weight data;

Based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logic contribution weight data and the knowledge point type text information set, a knowledge point contribution analysis is performed to obtain a plurality of the knowledge point contribution scores.

Preferably, the knowledge point contribution analysis is performed based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logical contribution weight data and the knowledge point type text information set to obtain a plurality of the knowledge point contribution scores, including:

Assume the logical contribution weight data is , the basic weight data of knowledge points is , the knowledge point type text information set is ,in is the number of knowledge point types, knowledge point type The set of distances from all argument connectives ,in To prove the number of conjunctions, Knowledge point type Connective words with argument , then for the first The knowledge point contribution score of each solution It is expressed as follows:

,

,

in Knowledge point type The basic weight value of Knowledge point type The contribution score of The basic weight and argument connective influence weight adjustment factor, default is 0.5.

Preferably, the knowledge point type segmentation processing is performed on the answer text to obtain multiple knowledge point type text information sets and multiple argumentation conjunctions, including:

Using a word segmentation tool to perform word segmentation processing on the answer text to obtain multiple answer word segmentation data;

Collect reference knowledge point word segmentation data containing L different knowledge point types, and obtain multiple reference argumentation connectives;

Integrate the plurality of the answer word segmentation data, the plurality of the reference knowledge point word segmentation data and the plurality of the reference argumentation connectives to obtain a mixed text data set;

Using K-means algorithm to perform L+1 cluster classification processing on the mixed text data set to obtain L+1 classification cluster data;

Analyze the distribution of the answer word segmentation data in the L+1 classification cluster data in turn, determine the answer word segmentation data in the classification cluster data containing the most reference argument connectives as argument connectives, and determine the knowledge point type of the answer word segmentation data as the knowledge point type corresponding to the reference knowledge point word segmentation data with the largest data volume in the classification cluster data.

Preferably, after collecting the test question data to be marked, the method further includes:

The text in the collected test data is preprocessed, including text normalization, mathematical formula conversion processing and mathematical entity synonymization processing.

Preferably, before acquiring the knowledge point basic weight data from the preset knowledge point type basic contribution weight table, the method further includes:

Constructing the basic contribution weight table of the knowledge point type, specifically including:

Collecting history test questions and solutions to obtain history test question data, and obtaining a teaching syllabus, wherein the teaching syllabus includes all knowledge point types;

Combined with the frequency of occurrence of the knowledge point type in the history test data, a statistical method is used to calculate the contribution weight of each knowledge point type.

Preferably, after obtaining the plurality of knowledge point contribution scores, the method further comprises:

The integrated knowledge point annotation data is visualized to generate a knowledge graph.

In a second aspect, the present invention provides a test question knowledge point annotation system based on artificial intelligence, the system comprising:

A data collection module is used to collect test data to be marked, wherein the test data includes a question text and a plurality of answer texts, each of which corresponds to a different problem-solving strategy;

A data processing module, used to perform knowledge point contribution analysis on the plurality of answer texts in sequence to obtain a plurality of knowledge point contribution scores;

The data integration module is used to integrate the contribution scores of the multiple knowledge points corresponding to each of the answer texts based on the question text to obtain multi-solution knowledge point annotation data.

Preferably, the data processing module includes:

The answer text segmentation unit is used to perform knowledge point type segmentation processing on the answer text to obtain multiple knowledge point type text information sets and multiple argumentation connectives;

A static data acquisition unit, used to acquire knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and based on a plurality of said argument connectives, match corresponding contribution weights from a preset argument connective contribution weight table to obtain logical contribution weight data;

The knowledge point contribution analysis unit is used to perform knowledge point contribution analysis based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logical contribution weight data and the knowledge point type text information set to obtain multiple knowledge point contribution scores.

Preferably, the system further comprises:

The visualization module is used to visualize the integrated knowledge point annotation data and generate a knowledge graph.

The beneficial effects of the present invention are:

(1) The present invention significantly improves the comprehensiveness of knowledge point annotation by introducing a multi-solution knowledge point annotation strategy. Traditional single-solution annotation is often limited to one thinking mode and easily ignores the multiple solutions to the problem and the potential knowledge point network. The present invention analyzes multiple problem-solving strategies and can not only capture the unique knowledge points behind different solutions, but also reveal the internal connections between knowledge points, providing students with a multi-angle learning perspective and promoting the cultivation of critical thinking and innovation ability.

(2) Based on the positional relationship between the basic weight data of knowledge points and argument connectives, combined with the logical contribution weight data and the knowledge point type text information set, the present invention realizes the refinement and intelligence of knowledge point contribution analysis. By quantifying the basic weights of knowledge points and the logical contributions of argument connectives, the actual role of knowledge points in the answering process can be accurately evaluated. It not only focuses on the static importance of knowledge points, but also considers their dynamic contributions in logical reasoning, thereby providing a more comprehensive and in-depth perspective for the evaluation of knowledge points.

(3) By introducing the influence weight and distance factor of argument connectives, the present invention can more accurately evaluate the role of knowledge points in the problem-solving process on the basis of quantitative analysis, and provide data-based decision support for educational decision makers and teachers. Educational decision makers can optimize course design and reasonably allocate educational resources based on the contribution scores and relevance of knowledge points; teachers can adjust teaching strategies based on the evaluation results of knowledge points to ensure that core knowledge points are fully explained and auxiliary knowledge points are appropriately covered, thereby improving teaching quality and efficiency.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the background technology, the drawings required for use in the embodiments of the present invention or the background technology will be described below.

The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and, together with the specification, are used to illustrate the technical solutions disclosed in the present invention.

FIG1 is a schematic diagram of a flow chart of a method for labeling test question knowledge points based on artificial intelligence provided by an embodiment of the present invention;

FIG2 is a schematic diagram of a flow chart of knowledge point contribution score analysis of a single solution provided by an embodiment of the present invention;

FIG3 is a schematic diagram of a flow chart of performing knowledge point type word segmentation processing on an answer text according to an embodiment of the present invention;

FIG4 is a schematic diagram of the structure of an artificial intelligence-based test question knowledge point annotation system provided in an embodiment of the present invention.

DETAILED DESCRIPTION

In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

The term "and/or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and/or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.

Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

In addition, in order to better illustrate the present invention, numerous specific details are provided in the following specific embodiments. It should be understood by those skilled in the art that the present invention can be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present invention.

At present, the existing test knowledge point annotation methods rely too much on a single solution, which leads to the one-sidedness and limitations of knowledge point annotation.

Example 1

The present invention provides a method for labeling test knowledge points based on artificial intelligence. Referring to FIG1 , the method includes:

S100, collecting test data to be annotated, the test data including a question text and a plurality of answer texts, each answer text corresponding to a different problem-solving strategy;

S200, performing knowledge point contribution analysis on multiple answer texts in sequence to obtain multiple knowledge point contribution scores;

S300, based on the question text, integrating the contribution scores of multiple knowledge points corresponding to each answer text to obtain multi-solution knowledge point annotation data.

In this embodiment, the test data to be annotated is collected from the question bank. In addition to the question text, each test question should also be accompanied by at least two different answer texts, and each answer text represents an independent problem-solving strategy. Furthermore, natural language processing (NLP) technology and machine learning algorithms are used to deeply analyze each answer text, identify and quantify the knowledge points involved, and calculate the contribution of each knowledge point to the answer text (i.e., the knowledge point contribution score). This process may involve semantic understanding, keyword extraction, concept mapping and other technologies. Finally, the knowledge point contribution scores of all answer texts under the same test question are summarized, and the comprehensive knowledge point annotation data of the test question is generated by weighted average or other statistical methods. In this way, even for the same question, the multi-dimensional knowledge point coverage brought by different problem-solving strategies can be reflected.

In one possible embodiment, test data is collected from an education database, an online education platform, or materials submitted by a teacher. Each test data contains at least one question text and multiple answer texts, ensuring that each answer text represents a different problem-solving strategy, removing noise information in the text, such as irrelevant symbols, spaces, etc., to ensure the accuracy of subsequent analysis, and using NLP technology to perform semantic parsing on each answer text to identify the knowledge points involved. This step may include techniques such as part-of-speech tagging, named entity recognition, and dependency syntax analysis to understand the core concepts and logical relationships in the answer text. The identified knowledge points are matched with a predefined knowledge point library to calculate a contribution score for each knowledge point. The calculation of the contribution score can be based on a comprehensive evaluation of multiple factors such as the frequency of occurrence of the knowledge point in the answer text, criticality, and relevance to the question. Considering that different problem-solving strategies may have different coverage of knowledge points, a weight is assigned to each answer text to reflect its importance in the overall knowledge point annotation. Based on the question text, the weighted average of the multiple knowledge point contribution scores corresponding to each answer text is performed to generate a comprehensive knowledge point annotation result. This process can use statistical methods, such as weighted average, maximum value selection, etc., to ensure the comprehensiveness and accuracy of the labeling results.

Compared with the traditional knowledge point annotation that only focuses on a single solution, the present invention can capture more knowledge points under more problem-solving paths, thereby providing more comprehensive learning resources, helping students understand and master different problem-solving ideas. Furthermore, multi-solution knowledge point annotation provides students with multiple angles for solving problems, which helps to stimulate students' innovative thinking and critical thinking ability, while also meeting the needs of students with different learning styles and levels, and promoting personalized learning. Teachers can use these multi-angle knowledge point annotations to design more targeted teaching activities, such as organizing discussions, case analysis, etc., to improve teaching effectiveness and student participation.

Preferably, referring to FIG2 , knowledge point contribution analysis is performed on multiple answer texts in sequence to obtain multiple knowledge point contribution scores, including:

S210, performing knowledge point type word segmentation processing on the answer text to obtain multiple knowledge point type text information sets and multiple argumentation connectives;

S220, obtaining knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and matching corresponding contribution weights from a preset argumentation connective contribution weight table based on a plurality of argumentation connectives to obtain logic contribution weight data;

S230, based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logic contribution weight data and the knowledge point type text information set, a knowledge point contribution analysis is performed to obtain a plurality of knowledge point contribution scores.

In this embodiment, a deep learning model (such as BiLSTM-CRF, BERT, etc.) is used for semantic understanding and word segmentation, to identify key words in the answer text, and to classify them into preset knowledge point types, such as mathematical formulas, theorem references, experimental steps, etc. At the same time, argumentation connectives in the text, such as "therefore", "however", "and", etc., are extracted for analyzing the logic of problem solving. The basic weight of each identified knowledge point type is searched and obtained from the pre-established knowledge point type basic contribution weight table. This table is based on the experience of domain experts and big data analysis, and reflects the importance of different knowledge points for problem solving. According to the extracted argumentation connectives, the corresponding logical contribution weights are matched from the preset argumentation connective contribution weight table (which reflects the logical value of different argumentation connectives in the problem solving process). This step takes into account the logical coherence in the problem solving process and ensures that the contribution of knowledge points in the argument chain is accurately quantified. Further, combined with the knowledge point basic weight data and the logical contribution weight data, according to the positional relationship between the knowledge point and the argumentation connective in the knowledge point type text information set, a comprehensive evaluation is performed through an algorithm (such as weighted summation, hierarchical analysis method, etc.) to obtain the specific contribution score of each knowledge point.

In this embodiment, the word segmentation technology based on deep learning and the preset weight table can more accurately identify and quantify the contribution of knowledge points, avoid the subjectivity and inconsistency of manual annotation, and improve the accuracy of knowledge point annotation. By analyzing and demonstrating the conjunctions and their contribution weights, the logical coherence of the knowledge point contribution analysis is enhanced, ensuring that the annotation results not only cover the knowledge points themselves, but also reflect the logical relationship between them, which is conducive to students' understanding of the internal logic in the problem-solving process. Furthermore, the automated knowledge point contribution analysis process greatly reduces the need for manual intervention and improves work efficiency. It also provides a feasible solution for the knowledge point annotation of large-scale test question banks and promotes the intelligent management of educational resources.

Preferably, based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logical contribution weight data and the knowledge point type text information set, a knowledge point contribution analysis is performed to obtain multiple knowledge point contribution scores, including:

Assume the logical contribution weight data is , the basic weight data of knowledge points is , the knowledge point type text information set is ,in is the number of knowledge point types, knowledge point type The set of distances from all argument connectives ,in To prove the number of conjunctions, Knowledge point type Connective words with argument , then for the first The knowledge point contribution score of each solution It is expressed as follows:

,

,

in Knowledge point type The basic weight value of Knowledge point type The contribution score of The basic weight and argument connective influence weight adjustment factor, default is 0.5.

In this embodiment, the logical contribution weight data Indicates the weight of each argument connective's contribution to the knowledge point, reflecting the impact of the argument structure on the value of the knowledge point, and the basic weight data of the knowledge point The basic importance index preset for each knowledge point type is not affected by the argument structure. is a moderating factor used to balance the contribution of the base weight and the influence of the argument connectives. By default, It is set to 0.5, which means that the basic weight is equally important as the weight of the argument connective influence, and can be adjusted according to actual conditions.

By introducing the influence weights of argument connectives and the distance factor, the present invention can more comprehensively evaluate the role of knowledge points in the answering process, avoiding the limitation of relying solely on basic weights, making the knowledge point contribution score more objective and comprehensive. Furthermore, the setting of the adjustment factor β allows the user to adjust the proportion of basic weights and the influence of argument connectives according to actual conditions, increasing the flexibility and adaptability of the model and making it suitable for test questions of different subject areas and complexity. The introduction of enables the model to finely analyze the positional relationship between knowledge points and argument connectives, and more accurately reflects the actual contribution of knowledge points in the solution logic. By quantifying the contribution of knowledge points in different solutions, teachers and students can gain in-depth insights into problem-solving strategies, promote the understanding and application of the essence of knowledge points, and thus improve the quality of education and learning outcomes.

Preferably, referring to FIG3 , the answer text is segmented by knowledge point type to obtain a plurality of knowledge point type text information sets and a plurality of argumentation conjunctions, including:

S211, using a word segmentation tool to perform word segmentation processing on the answer text to obtain a plurality of answer word segmentation data;

S212, collecting reference knowledge point segmentation data including L different knowledge point types, and obtaining a plurality of reference argumentation connectives;

S213, integrating a plurality of answer segmentation data, a plurality of reference knowledge point segmentation data, and a plurality of reference argumentation connectives to obtain a mixed text data set;

S214, using K-means algorithm to perform L+1 cluster classification processing on the mixed text data set to obtain L+1 classification cluster data;

S215, analyze the distribution of the answer word segmentation data in the L+1 classification cluster data in turn, determine the answer word segmentation data in the classification cluster data containing the most reference argument connectives as argument connectives, and determine the knowledge point type of the answer word segmentation data as the knowledge point type corresponding to the reference knowledge point word segmentation data with the largest data volume in the classification cluster data.

In this embodiment, advanced natural language processing (NLP) word segmentation tools, such as Jieba word segmentation, StanfordNLP, etc., are used to perform preliminary vocabulary segmentation on the answer text to obtain a series of answer word segmentation data. Reference texts containing L different types of knowledge points are collected and word segmentation is performed on them to obtain reference knowledge point word segmentation data, which represent typical expressions of known knowledge point types, and a list of multiple common argumentation conjunctions, such as "therefore", "however", "in addition", etc., are sorted out to identify the logical structure in the answer text, and the answer word segmentation data, reference knowledge point word segmentation data and reference argumentation conjunctions are merged to form a comprehensive mixed text data set. Further, the K-means algorithm is used to classify the mixed text data set into L+1 clusters, where L is the number of reference knowledge point types, and an additional cluster is used to accommodate knowledge points or argumentation conjunctions that are not clearly attributed. The K-means algorithm classifies similar words into the same cluster through iterative calculations to form L+1 classified cluster data. Then, the distribution of the answer segmentation data in the L+1 classification cluster data is analyzed one by one, and the classification cluster containing the most reference argumentation connectives is identified and marked as the argumentation connective cluster. For other classification clusters, the knowledge point type corresponding to the reference knowledge point segmentation data with the largest amount of data is determined as the knowledge point type of the answer segmentation data in the cluster.

In this embodiment, by means of the K-means clustering algorithm, the method can effectively distinguish different types of knowledge points and argument connectives, improve the accuracy of recognition, and avoid the subjectivity and errors of manual judgment. Furthermore, since the reference knowledge point segmentation data and argument connectives are used as classification references, the method has strong adaptability and can be applied to different disciplines, such as mathematics, physics, chemistry, etc., and only the reference knowledge points need to be adjusted accordingly.

Preferably, after collecting the test question data to be marked, the method further includes:

The text in the collected test data is preprocessed, including text normalization, mathematical formula conversion processing and mathematical entity synonymization processing.

In this embodiment, the text content is converted into a unified encoding format, such as UTF-8, to ensure the correct display and processing of the text between different systems and devices. Useful characters, blanks, punctuation marks, etc. in the text are removed by regular expressions or NLP tools to maintain the cleanliness of the text and reduce interference in subsequent analysis. Correct spelling errors, unify the writing form of abbreviations and terms, and ensure the consistency and professionalism of the text. Use specialized mathematical formula recognition software or tools, such as Mathpix, LaTeXML, etc., to accurately identify and extract mathematical formulas from the text. Convert the extracted mathematical formulas to a standard format, such as MathML or LaTeX, for easy computer understanding and processing. This process may involve parsing and reconstruction of the formula structure. The converted mathematical formula is re-embedded into the original text to ensure the accuracy of its position and context. Use NLP technology to identify mathematical entities in the text, such as variable names, function names, constants, etc., establish a synonym library of mathematical entities, and include mathematical entities with different expressions but the same or similar meanings, such as π and pi, sin and sine function, etc. According to the synonym library, replace the mathematical entities appearing in the text with their standard or unified expression forms to eliminate the understanding barriers caused by differences in expression.

In this embodiment, text normalization ensures data consistency and readability, mathematical formula conversion processing ensures formula accuracy and processability, and mathematical entity synonymization processing eliminates expression differences, which together improve the overall quality of the data. High-quality preprocessing results provide a solid foundation for subsequent knowledge point identification and contribution analysis, reduce analysis deviations caused by data quality issues, and enhance the accuracy of the final annotation results. The standardized and normalized data format simplifies subsequent processing procedures, reduces unnecessary data cleaning work, speeds up the entire analysis process, and improves work efficiency.

Preferably, before obtaining the knowledge point basic weight data from the preset knowledge point type basic contribution weight table, the method further includes:

Construct a basic contribution weight table of knowledge point types, including:

Collect history test questions and solutions to obtain history test data, and obtain the teaching syllabus, which contains all types of knowledge points;

Based on the frequency of occurrence of knowledge point types in historical test data, statistical methods are used to calculate the contribution weight of each knowledge point type.

In this embodiment, the examination questions of previous years and their official or recognized solutions are collected from channels such as education departments, schools, and examination institutions to ensure the authority and reliability of the data. The collected historical examination question data are classified and sorted, and divided according to the dimensions of subject, grade, examination type, etc., to facilitate subsequent analysis and processing. Further, the latest teaching syllabus is obtained from the education department or the official textbook publishing unit to ensure that all necessary knowledge point types are covered. The teaching syllabus is parsed, all knowledge point types are extracted, and a list of knowledge point types is formed to provide a reference for subsequent weight calculation. Based on the historical examination question data, text analysis and data mining techniques are used to count the number of occurrences of each knowledge point type in the examination question, reflecting the importance and commonness of the knowledge point in the examination. Statistical methods such as frequency analysis, principal component analysis, regression analysis, etc. are used to combine the frequency of occurrence of the knowledge point type in the historical examination question to calculate the contribution weight of each knowledge point type. This process may also need to consider factors such as the difficulty and relevance of the knowledge point. According to the requirements of the teaching syllabus and the opinions of education experts, the calculated contribution weight is manually calibrated to ensure the rationality and educational applicability of the weight.

In this embodiment, based on the historical test data and teaching syllabus, the contribution weights are calculated using statistical methods to ensure the scientificity and objectivity of the weight setting and avoid the influence of human bias. Furthermore, the construction of the weight table is closely aligned with the teaching syllabus, ensuring the consistency of knowledge point annotations with educational objectives, helping to guide students and teachers to focus on the key points and difficulties of the exam, and improving teaching efficiency and learning outcomes. Furthermore, by regularly updating the historical test data and teaching syllabus, the knowledge point type basic contribution weight table can be adjusted in a timely manner to reflect the latest educational trends and exam requirements, and to maintain the timeliness and effectiveness of the weight table.

Preferably, after obtaining the contribution scores of the multiple knowledge points, the method further includes:

The integrated knowledge point annotation data is visualized to generate a knowledge graph.

In this embodiment, the quality check of the integrated knowledge point annotation data is performed to remove duplicate, missing or abnormal data points to ensure the accuracy and consistency of the data. The knowledge point contribution score is standardized, for example, Z-score standardization or Min-Max normalization is used so that data of different dimensions can be compared on the same scale. Data mining techniques, such as Apriori algorithm or FP-growth algorithm, are used to discover the association rules between knowledge points, identify which knowledge points often appear together in the same test questions, and reflect the potential connection between knowledge points. Using algorithms such as K-means, hierarchical clustering or DBSCAN, knowledge points are divided into different subject areas according to the knowledge point contribution scores and association rules to reveal the knowledge point structure. The knowledge points are used as nodes in the knowledge graph, and each node contains the name, description, contribution score and other attributes of the knowledge point. According to the association rules between knowledge points and the clustering results of the subject area, the edges between nodes are defined to represent the degree of association and logical relationship between knowledge points. Choose a suitable graph layout algorithm, such as the Fruchterman-Reingold algorithm or the Spring-Electric model, to determine the position of nodes in two-dimensional or three-dimensional space, so that the structure of the knowledge graph is clear and easy to understand. Design different colors, shapes, and sizes for nodes and edges to distinguish different types of knowledge points and association strengths, and enhance visual expression. Develop functions such as hovering the mouse to display detailed information, clicking to zoom in to view details, and dragging to re-layout to enhance the user experience of exploring the knowledge graph.

In this embodiment, the knowledge graph graphically displays the complex relationship between knowledge points, helping users quickly grasp the overall structure and mutual influence of knowledge points, and promoting understanding and memory of the knowledge system. Furthermore, teachers can identify teaching key points and difficulties based on the knowledge graph, adjust course content and teaching strategies; students can discover their own knowledge blind spots through the knowledge graph and supplement their learning in a targeted manner.

Example 2

The present invention provides a test question knowledge point annotation system based on artificial intelligence. Referring to FIG4 , the system includes:

The data collection module 100 is used to collect the test data to be marked, the test data includes the question text and multiple answer texts, each answer text corresponds to a different problem-solving strategy;

The data processing module 200 is used to perform knowledge point contribution analysis on multiple answer texts in sequence to obtain multiple knowledge point contribution scores;

The data integration module 300 is used to integrate the multiple knowledge point contribution scores corresponding to each answer text based on the question text to obtain multi-solution knowledge point annotation data.

Preferably, the data processing module includes:

The answer text segmentation unit is used to perform knowledge point type segmentation processing on the answer text to obtain multiple knowledge point type text information sets and multiple argumentation connectors;

A static data acquisition unit is used to acquire knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and based on a plurality of argument connectives, matches corresponding contribution weights from a preset argument connective contribution weight table to obtain logical contribution weight data;

The knowledge point contribution analysis unit is used to perform knowledge point contribution analysis based on the knowledge point basic weight data and the positional relationship between the knowledge point and the argument connective, and according to the logical contribution weight data and the knowledge point type text information set to obtain multiple knowledge point contribution scores.

Preferably, with continued reference to FIG. 4 , the system further includes:

The visualization module 400 is used to visualize the integrated knowledge point annotation data and generate a knowledge graph.

The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims (9)

1. The method for marking the test question knowledge points based on the artificial intelligence is characterized by comprising the following steps of:

collecting test question data to be marked, wherein the test question data comprises a question text and a plurality of answer texts, and each answer text corresponds to different answer strategies;

sequentially carrying out knowledge point contribution analysis processing on a plurality of answer texts to obtain a plurality of knowledge point contribution scores;

integrating a plurality of knowledge point contribution scores corresponding to each answer text based on the topic text to obtain multi-solution knowledge point labeling data;

the step of sequentially carrying out knowledge point contribution analysis processing on a plurality of answer texts to obtain a plurality of knowledge point contribution scores comprises the following steps:

Carrying out knowledge point type word segmentation processing on the answer text to obtain a plurality of knowledge point type text information sets and a plurality of demonstration connecting words;

Acquiring knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and matching corresponding contribution weights from the preset proof connecting word contribution weight table based on a plurality of proof connecting words to obtain logic contribution weight data;

And carrying out knowledge point contribution analysis according to the logic contribution weight data and the knowledge point type text information set based on the knowledge point basic weight data and the position relation between the knowledge points and the demonstration connective, so as to obtain a plurality of knowledge point contribution scores.

2. The method for labeling test question knowledge points based on artificial intelligence according to claim 1, wherein the steps of obtaining a plurality of knowledge point contribution scores based on the knowledge point basic weight data and the positional relationship between knowledge points and arguments, and performing knowledge point contribution analysis according to the logic contribution weight data and the knowledge point type text information set, comprise:

Let the logic contribution weight data be The knowledge point basic weight data isKnowledge point type text information set isWhereinFor the number of knowledge point types, the knowledge point typesDistance set to all argumentsWhereinIn order to demonstrate the number of connective words,For knowledge point typeAnd arguments of conjunctionsFor the distance of (1)Knowledge point contribution score for a personalized solutionThe expression is as follows:

，

，

Wherein the method comprises the steps of For knowledge point typeIs used as a basis weight value for the (c),For knowledge point typeIs used to determine the contribution score of (c),Influencing the weight adjustment factor for the basic weight and the arguments, defaulting0.5.

3. The method for labeling question knowledge points based on artificial intelligence according to claim 1, wherein the step of performing knowledge point type word segmentation on the answer text to obtain a plurality of knowledge point type text information sets and a plurality of argumentation connective words comprises the steps of:

Performing word segmentation processing on the answer text by using a word segmentation tool to obtain a plurality of answer word segmentation data;

collecting word segmentation data of reference knowledge points containing L different knowledge point types, and obtaining a plurality of reference demonstration connecting words;

Integrating a plurality of the solution word segmentation data, a plurality of the reference knowledge point word segmentation data and a plurality of the reference proof connecting words to obtain a mixed text data set;

carrying out L+1 cluster classification processing on the mixed text data set by adopting a K-means algorithm to obtain L+1 classification cluster data;

And sequentially analyzing the distribution condition of the solution word segmentation data in the L+1 classification cluster data, determining the solution word segmentation data in the classification cluster data containing the maximum reference demonstration connecting words as the demonstration connecting words, and determining the knowledge point type of the solution word segmentation data as the knowledge point type corresponding to the reference knowledge point word segmentation data with the maximum data quantity in the classification cluster data.

4. The method for labeling test question knowledge points based on artificial intelligence according to claim 1, wherein after collecting test question data to be labeled, further comprises:

And preprocessing the text in the acquired test question data, wherein the preprocessing comprises text normalization, mathematical formula conversion processing and mathematical entity synonymous processing.

5. The method for labeling test question knowledge points based on artificial intelligence according to claim 1, wherein before the knowledge point basic weight data is obtained from a preset knowledge point type basic contribution weight table, the method further comprises:

constructing the knowledge point type basic contribution weight table specifically comprises the following steps:

Collecting historical examination questions and a solution to obtain historical examination question data, and obtaining a teaching outline, wherein the teaching outline comprises all knowledge point types;

And combining the occurrence frequency of the knowledge point types in the historical test question data, and calculating the contribution weight of each knowledge point type by adopting a statistical method.

6. The method for labeling question knowledge points based on artificial intelligence according to claim 1, wherein after obtaining the plurality of knowledge point contribution scores, further comprises:

And visually displaying the integrated knowledge point labeling data to generate a knowledge graph.

7. An artificial intelligence based test question knowledge point labeling system applied to the test question knowledge point labeling method based on the artificial intelligence as claimed in any one of claims 1 to 6, wherein the system comprises:

The data acquisition module is used for acquiring test question data to be marked, wherein the test question data comprises a question text and a plurality of answer texts, and each answer text corresponds to different answer strategies;

The data processing module is used for sequentially carrying out knowledge point contribution analysis processing on the multiple answer texts to obtain multiple knowledge point contribution scores;

and the data integration module is used for integrating a plurality of knowledge point contribution scores corresponding to each answer text based on the topic text to obtain multi-solution knowledge point labeling data.

8. The system for labeling test question knowledge points based on artificial intelligence according to claim 7, wherein the data processing module comprises:

The answering text word segmentation unit is used for carrying out knowledge point type word segmentation processing on the answering text to obtain a plurality of knowledge point type text information sets and a plurality of demonstration connecting words;

the static data acquisition unit is used for acquiring knowledge point basic weight data from a preset knowledge point type basic contribution weight table, and matching corresponding contribution weights from the preset argument connective contribution weight table based on a plurality of argument connective words to acquire logic contribution weight data;

And the knowledge point contribution analysis unit is used for carrying out knowledge point contribution analysis according to the knowledge point basic weight data and the position relation between the knowledge points and the demonstration connecting words and the logic contribution weight data and the knowledge point type text information set to obtain a plurality of knowledge point contribution scores.

9. The artificial intelligence based test question knowledge point labeling system of claim 7, further comprising:

And the visualization module is used for carrying out visual display on the integrated knowledge point labeling data to generate a knowledge graph.