An online management method for training

Abstract

The present invention provides an online management method for training, which belongs to the technical field of training management, and includes: step 1: establishing a knowledge graph based on preset learning content; step 2: performing a first evaluation on the starting level and learning needs of the trainees, and constructing the learning path of the trainees according to the first evaluation result and the knowledge graph; step 3: determining a number of corresponding management threads based on the learning path of each trainee, and monitoring the learning activities of the corresponding trainees in real time based on the management threads and preset monitoring tools; step 4: performing a second evaluation on the learning activities of the corresponding trainees, and adjusting the training strategy according to the second evaluation result and performing trainee training management. Through the automation and personalized management of the entire training process, the efficiency of training management is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of training management, and in particular to an online management method for training. Background Art

[0002] With the development of information technology and the popularization of the Internet, people's demand for online learning is growing. Traditional education and training methods can no longer meet the needs of modern society for flexibility, personalization and efficiency. Therefore, online education and training has become a trend, and various online learning platforms and training management systems have been developed and applied.

[0003] Among the existing technical solutions, similar methods usually adopt personalized learning path generation based on learner data and content analysis, using machine learning algorithms and big data analysis to generate personalized learning paths based on learners' previous performance and learning goals. Although the existing technical solutions have achieved the generation of personalized learning paths and the recommendation of learning resources, there are still some technical defects in real-time monitoring of learning activities and dynamic adjustment of training strategies.

[0004] Therefore, the present invention provides an online management method for training. Summary of the invention

[0005] The present invention provides an online management method for training, which is used to perform a first assessment on the starting level and learning needs of trainees, and to build a personalized learning path according to a knowledge graph, to monitor the learning activities of trainees in real time based on management threads and preset monitoring tools, and to adjust the training strategy in time, and to perform a second assessment on the learning activities of trainees and to adjust the training strategy according to the assessment results, so that the training strategy can be optimized in time according to actual conditions, thereby improving the efficiency and effectiveness of training.

[0006] The present invention provides an online management method for training, comprising:

[0007] Step 1: Create a knowledge graph based on the preset learning content;

[0008] Step 2: Conduct a first assessment of the student's starting level and learning needs, and construct a learning path for the student based on the first assessment result and the knowledge graph;

[0009] Step 3: Determine a number of corresponding management threads based on the learning path of each student, and monitor the learning activities of the corresponding students in real time based on the management threads and preset monitoring tools;

[0010] Step 4: Conduct a second assessment of the learning activities of the corresponding trainees, adjust the training strategy and conduct trainee training management based on the results of the second assessment.

[0011] The present invention provides an online management method for training, which establishes a knowledge graph based on preset learning content, including:

[0012] Extract knowledge text from preset learning content based on natural language processing technology;

[0013] Decompose the knowledge text into content and establish the knowledge unit structure;

[0014] Build a knowledge graph based on the knowledge unit structure and preset layout.

[0015] The present invention provides an online management method for training, which decomposes the content of a knowledge text to establish a knowledge unit structure, including:

[0016] Segment the knowledge text based on preset segmentation rules;

[0017] Determine the number of windows according to the number of segments of the knowledge text, the segment length of each segment text, and the number of text sentences involved in the segment text;

[0018] Dividing the segmented text according to the number of windows to obtain the last matching segment of each window, and resizing the window to place the corresponding last matching segment;

[0019] Based on all the text sentences involved in each window, construct the vector of the corresponding window;

[0020] Decompose the vector of each window and extract the key features of each window;

[0021] Determine the core theme of each window based on its key features;

[0022] Treat each core topic as a knowledge unit and determine the nodes of the knowledge unit structure based on each knowledge unit;

[0023] Conduct dependency analysis on each core topic to determine the dependencies between different topics;

[0024] Determine the connection relationship between nodes based on the dependencies between different topics;

[0025] The knowledge unit structure is constructed based on all nodes and the connection relationships between nodes.

[0026] The present invention provides an online management method for training, wherein preset segmentation rules include: paragraph segmentation, punctuation segmentation, keyword segmentation, word count segmentation and theme conversion segmentation.

[0027] The present invention provides an online management method for training, which adjusts the size of the window to place the corresponding last matching segment, including:

[0028] Determine the sentence length of each text sentence in the last matching segment of each window, and at the same time, determine the initial size of each window;

[0029] The initial size is adjusted according to the sentence length to obtain the adjusted window size:

[0030] ;in, is the adjusted window size of the i-th window, is the average sentence length of all text sentences in the i-th window, is the length of the qth segment text to be placed in the i-th window, is the average segment length of all segmented texts required to be placed in the i-th window, n1 is the number of segments required to be placed in the i-th window, is the influence control factor of the i-th window on the q-th segment text, is the preset weight coefficient corresponding to the average sentence length in the i-th window, is the preset weight coefficient corresponding to the average segment length under the i-th window, is the initial size of the i-th window, is the average number of sentences involved in each segment text under the i-th window, For all The minimum length in .

[0031] The present invention provides an online management method for training, which constructs a vector of a corresponding window based on all text sentences involved in each window, including:

[0032] Extract the vocabulary of the text sentences involved in each window;

[0033] Creating a vocabulary index for each vocabulary based on a preset vocabulary-index data table, and then building a vocabulary table corresponding to each sentence based on a preset table building method;

[0034] Calculate the TF-IDF value for each word:

[0035] ;in, is the TF-IDF value of the jth word in the dth text sentence in the ith window, is the preset word frequency control coefficient of the jth word in the dth text sentence in the ith window, and , is the frequency of occurrence of the jth word in the dth text sentence in the ith window, is the word with the highest frequency in the dth text sentence in the ith window The frequency of occurrence, is the number of text sentences in the ith window that contain the same word as the jth word in the dth text sentence, is the total number of text sentences in the i-th window, is the preset word weight of the jth word in the dth text sentence in the ith window, is the length of the dth text sentence in the ith window;

[0036] The TF-IDF value of each word in each window is filled into the vocabulary in turn to obtain the sentence vector of the corresponding text sentence, and then the vector of the corresponding window is obtained.

[0037] The present invention provides an online management method for training, which constructs a learning path for the trainee according to a first evaluation result and a knowledge graph, including:

[0038] Based on the first assessment result, determining the starting level and learning needs of the student and several key features associated with the knowledge graph;

[0039] Mapping each associated key feature into the state space to obtain the estimated learning state of the learner based on each knowledge unit structure;

[0040] Determining, based on the estimated learning state, an optional learning action for the user based on the corresponding knowledge unit structure;

[0041] Determine the action selection strategy based on the optional learning actions and the preset habit function;

[0042] According to the user action selection strategy, effective learning actions are selected from the optional learning actions of each knowledge unit structure, and based on the sequence of the knowledge unit structures in the knowledge graph, the learning path of the student is obtained.

[0043] The present invention provides an online management method for training, mapping each associated key feature into a state space, including:

[0044] Determine a value range of each associated key feature based on the feature type corresponding to each associated key feature;

[0045] Defining the dimensions of the student's state space based on each associated key feature and the corresponding value range, wherein each dimension corresponds to a key feature;

[0046] Combining the value of each associated key feature with the value of each other associated key feature based on a preset combination method to determine all possible state combinations;

[0047] The state space is constructed based on all possible state combinations.

[0048] The present invention provides an online management method for training, which performs a first assessment on the starting level and learning needs of trainees, builds a personalized learning path based on a knowledge graph, monitors the learning activities of trainees in real time based on management threads and preset monitoring tools, and adjusts the training strategy in time, performs a second assessment on the learning activities of trainees, and adjusts the training strategy according to the assessment results, so that the training strategy can be optimized in time according to actual conditions, thereby improving the efficiency and effectiveness of training. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a flow chart of an online management method for training provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0052] Example 1

[0053] like Figure 1 As shown, the present invention provides an online management method for training, comprising:

[0054] Step 1: Create a knowledge graph based on the preset learning content;

[0055] Step 2: Conduct a first assessment of the student's starting level and learning needs, and construct a learning path for the student based on the first assessment result and the knowledge graph;

[0056] Step 3: Determine a number of corresponding management threads based on the learning path of each student, and monitor the learning activities of the corresponding students in real time based on the management threads and preset monitoring tools;

[0057] Step 4: Conduct a second assessment of the learning activities of the corresponding trainees, adjust the training strategy and conduct trainee training management based on the results of the second assessment.

[0058] In this embodiment, the knowledge graph describes the organizational structure and association relationship of knowledge through the relationship between nodes and edges, and is used to describe the association and hierarchy between various learning contents, helping to determine the learning path and learning goals. For example, in a programming course, the knowledge graph can organize and associate various programming languages, algorithms, data structures and other knowledge points to form a complete knowledge system. For an English training course, the knowledge graph may include different knowledge points such as grammar, vocabulary, listening, and speaking, among which grammatical knowledge includes subject, predicate, object, tense, etc.; vocabulary knowledge includes commonly used words, phrases, etc.; listening knowledge includes listening skills, common scene dialogues, etc.

[0059] In this embodiment, the starting level of a student refers to the level of knowledge and skills that the student has mastered in a certain field or subject. For example, for an English training course, the starting level of a student can be assessed through a language proficiency test or previous learning experience to determine whether he is a beginner, intermediate student or advanced student. If a student already has certain English listening and speaking skills but is weak in grammar and vocabulary, then his starting level can be positioned as an intermediate student who needs to focus on strengthening his grammar and vocabulary knowledge;

[0060] In this embodiment, learning needs refer to the knowledge, skills or abilities that students need to improve in a certain field or topic. For example, one student may need to improve his or her English speaking ability to cope with communication needs at work, while another student may focus more on English writing ability to prepare for academic exams.

[0061] In this embodiment, the first assessment refers to a preliminary assessment of the student's starting level and learning needs, which is intended to determine the student's learning path and training content, and is conducted through tests, questionnaires, interviews, etc. to obtain the student's basic information and learning needs;

[0062] In this embodiment, the learning path is a personalized learning roadmap constructed according to the student's starting level and learning needs, guiding the student to learn in a certain order and steps. The learning path can be determined according to the knowledge points and learning objectives in the knowledge graph to ensure that the student systematically masters the necessary knowledge and skills during the training process. For example, a student's learning path may include learning basic grammar knowledge first, and then gradually expanding to vocabulary accumulation, listening training, oral practice, etc., to ultimately achieve the goal of improving comprehensive English ability.

[0063] In this embodiment, the management thread refers to a series of management and guidance measures determined according to the student's learning path, which are used to guide the student to conduct effective learning and training. The management thread includes course setting, learning task assignment, learning progress tracking, learning feedback, etc. For example, the management thread of an English training course may include activities such as weekly release of learning tasks, regular oral practice, participation in online discussions, as well as monitoring the student's learning progress and providing feedback on learning;

[0064] In this embodiment, the preset monitoring tool is a tool or system for real-time monitoring of students' learning activities, which can help teachers or administrators to promptly identify students' learning difficulties and problems, and provide necessary support and guidance. For example, a preset monitoring tool can be a learning data analysis module in a learning management system, which can automatically generate learning reports based on students' learning behaviors and academic performance, and promptly identify students' learning problems so that teachers and administrators can provide targeted guidance and help.

[0065] In this embodiment, the second assessment refers to an assessment of the learners' learning activities based on the learning path, which is intended to assess the learners' learning outcomes and learning effects, and adjust the training strategies and content according to the assessment results.

[0066] The working principle and beneficial effects of the above technical solution are: by conducting a first assessment of the starting level and learning needs of the trainees, and building a personalized learning path based on the knowledge graph, the trainees' learning activities are monitored in real time based on the management thread and the preset monitoring tool, and the training strategy is adjusted in time; by conducting a second assessment of the trainees' learning activities, and adjusting the training strategy according to the assessment results, the training strategy can be optimized in time according to the actual situation, thereby improving the efficiency and effectiveness of the training.

[0067] Example 2

[0068] The present invention provides an online management method for training, which establishes a knowledge graph based on preset learning content, including:

[0069] Extract knowledge text from preset learning content based on natural language processing technology;

[0070] Decompose the knowledge text into content and establish the knowledge unit structure;

[0071] Build a knowledge graph based on the knowledge unit structure and preset layout.

[0072] In this embodiment, natural language processing is used to extract knowledge text from preset learning content, that is, to analyze and process the text through algorithms and models to obtain useful information therein. For example, on an online education platform, natural language processing technology is used to extract key concepts, definitions, examples and other knowledge texts from learning resources such as textbooks and course outlines to construct a knowledge graph;

[0073] In this embodiment, knowledge text refers to text content containing useful knowledge, which can be text descriptions in learning resources such as textbooks, course outlines, reference materials, etc. Natural language processing technology is used to extract knowledge text from preset learning content to construct a knowledge graph. For example, if a training course is about programming, the knowledge text may include the grammatical rules of the programming language, descriptions of common algorithms, programming examples, etc.

[0074] In this embodiment, content decomposition is the process of breaking down complex knowledge or content into smaller, more manageable parts. The knowledge text is broken down into smaller knowledge unit structures to facilitate the construction of knowledge graphs and the management of learning content. For example, if a training course involves multiple topics, such as a mathematics course including algebra, geometry, calculus, etc., then each topic can be further broken down into individual knowledge points, such as equations, inequalities, functions, etc. in the algebra topic;

[0075] In this embodiment, the knowledge unit structure refers to the relationship and organizational structure between knowledge units obtained by content decomposition, which is used to construct a knowledge graph to represent the hierarchical relationship and mutual association between different knowledge units. For example, in a programming course, the knowledge unit structure can organize and associate different programming languages, algorithms, data structures and other knowledge points according to certain logical relationships.

[0076] In this embodiment, the preset layout refers to the graph structure and layout method predetermined when establishing the knowledge graph. The knowledge graph is constructed based on the knowledge unit structure and the preset layout, that is, the content of the knowledge graph is organized and presented according to pre-set rules and layout methods. For example, if the preset layout of a training course is to arrange the knowledge points in a logical order of learning, then the layout of the knowledge graph may be presented in order from basic to advanced, and from simple to complex.

[0077] The working principle and beneficial effects of the above technical solution are: by utilizing natural language processing technology, the system can automatically extract useful knowledge texts from preset learning content, lay the foundation for subsequent knowledge graph construction, decompose the extracted knowledge texts, and split complex learning content into smaller and more manageable knowledge units, making the knowledge graph clearer and easier to understand. Based on the obtained knowledge unit structure and preset layout, a knowledge graph is constructed, which helps to improve the efficiency and quality of management training.

[0078] Example 3

[0079] The present invention provides an online management method for training, which decomposes the content of a knowledge text to establish a knowledge unit structure, including:

[0080] Segment the knowledge text based on preset segmentation rules;

[0081] Determine the number of windows according to the number of segments of the knowledge text, the segment length of each segment text, and the number of text sentences involved in the segment text;

[0082] Dividing the segmented text according to the number of windows to obtain the last matching segment of each window, and resizing the window to place the corresponding last matching segment;

[0083] Based on all the text sentences involved in each window, construct the vector of the corresponding window;

[0084] Decompose the vector of each window and extract the key features of each window;

[0085] Determine the core theme of each window based on its key features;

[0086] Treat each core topic as a knowledge unit and determine the nodes of the knowledge unit structure based on each knowledge unit;

[0087] Conduct dependency analysis on each core topic to determine the dependencies between different topics;

[0088] Determine the connection relationship between nodes based on the dependencies between different topics;

[0089] The knowledge unit structure is constructed based on all nodes and the connection relationships between nodes.

[0090] In this embodiment, a window refers to an area where the knowledge text is divided, which is used to extract text sentences and construct vector representations. The size and number of windows are determined according to preset rules and are used to segment and analyze the knowledge text. For example, an article is divided into multiple windows, each window contains a certain number of text sentences, then each window is an independent area for subsequent processing and analysis.

[0091] In this embodiment, the last matching segment is the paragraph that is divided and included in each window during the window division process. For example, if a window includes the last few sentences of an article, then this text is the last matching segment of the window.

[0092] In this embodiment, the vector of the window is a mathematical vector representing the text sentences in the window, which is used to characterize the content features of the window. Each window can be represented as a vector, which contains the feature information of each text sentence in the window. The dimension of each vector corresponds to a word in a vocabulary, and each sentence represents a point in a high-dimensional space. Assume that the vocabulary is (["apple", "banana", "cherry"]). The sentence content is "apple banana apple", and the TF-IDF vector may be ([0.6, 0.3, 0]), where 0.6 and 0.3 represent the TF-IDF values ​​of "apple" and "banana" respectively, and "cherry" is 0 because it does not appear in the document. For example, a window contains multiple text sentences, and these text sentences can be converted into vector representations using technologies such as bag of words model or Word2Vec for subsequent analysis and processing.

[0093] In this embodiment, performing spectral decomposition on the vector of each window refers to extracting features from the vector of the window to obtain key features of the window. Spectral decomposition is a commonly used matrix decomposition method that can decompose a matrix into the form of eigenvalues ​​and eigenvectors to extract key features of the matrix. For example, by performing spectral decomposition on the vector of the window, the main features of the window, such as the subject or keywords contained in the window, can be extracted.

[0094] In this embodiment, the core topic is the main content or key topic of the window determined on the basis of the window, and is used to represent the core concepts or key contents involved in the window. For example, if a window mainly discusses the principles and applications of machine learning algorithms, then the machine learning algorithm is the core topic of the window.

[0095] In this embodiment, the knowledge unit is the basic unit for constructing the knowledge unit structure, representing a node in the knowledge graph, and contains a specific topic or concept. Each core topic is regarded as a knowledge unit and is used to construct the knowledge unit structure. For example, the knowledge graph involves multiple core concepts in the field of machine learning, such as regression analysis, classification algorithms, clustering methods, etc., then each core concept is a knowledge unit;

[0096] In this embodiment, dependency analysis refers to analyzing the dependency between different topics or knowledge units to determine the degree of association and mutual influence between them. Dependency analysis helps to understand the logical relationship and hierarchy between knowledge units. For example, in a programming course, the concept of function may depend on the concept of variable and control flow, so the dependency between function, variable and control flow needs to be analyzed when constructing the knowledge unit structure.

[0097] In this embodiment, the connection relationship refers to the connection relationship between different knowledge units in the knowledge unit structure, reflecting the logical relationship and mutual connection between them. The connection relationship can be direct or indirect, and is used to construct the edges of the knowledge unit structure. For example, in the knowledge graph, there is a direct association relationship between functions, variables and control flows, so there will be a connection relationship between them, reflecting the logical connection between them.

[0098] In this embodiment, the knowledge unit structure is a graphical structure composed of nodes and connection relationships between knowledge units, which is used to represent the hierarchical relationship and mutual correlation between different knowledge units in the knowledge graph. For example, the knowledge unit structure can organize and represent various knowledge units in a programming course, such as functions, variables, control flows, etc., according to certain logical relationships and hierarchical structures to facilitate students' understanding and learning.

[0099] The working principle and beneficial effects of the above technical solution are: the knowledge text is segmented according to the preset segmentation rules, and the number of windows is determined according to the number of segments, the length of each text and the number of sentences, which helps to decompose large text into more manageable parts, and determine the number of windows for subsequent processing, extract the text sentences in each window, and construct a vector representation of the corresponding window, decompose the vector of each window, and extract key features to determine the core theme of each window, which provides a basis for the construction of knowledge units, and take each core theme as a knowledge unit, and determine the nodes of the knowledge unit structure based on them. Then, dependency analysis is performed to determine the dependencies between different topics, thereby constructing the knowledge unit structure, making the logical relationship between knowledge units clearer, and determining the connection relationship between nodes according to the dependencies between different topics, which helps to construct a graphical representation of the knowledge unit structure, making the association relationship between knowledge units more intuitive and clear, and improving the efficiency and quality of training and learning.

[0100] Example 4

[0101] The present invention provides an online management method for training, wherein preset segmentation rules include: paragraph segmentation, punctuation segmentation, keyword segmentation, word count segmentation, and theme conversion segmentation.

[0102] In this embodiment, the paragraph segmentation rule is to segment based on the logical relationship between paragraphs and the topic conversion. For example, when a paragraph discusses a topic and ends, it can be regarded as a paragraph. For example, in an article about solar energy, each paragraph may introduce the principle, application, advantages and disadvantages of solar energy respectively;

[0103] In this embodiment, the punctuation segmentation rule is based on the semantic association between sentences. Common punctuation marks such as period, question mark, exclamation mark, etc. can be used as the basis for segmentation. For example, in a tutorial article, each sentence can be used as a paragraph, such as one sentence explaining a concept or a step;

[0104] In this embodiment, the keyword segmentation rule is to segment based on the occurrence or semantic mutation of keywords, which is usually related to the occurrence of key topics or important content. For example, in a scientific report, when a new technology or new discovery is mentioned, it can be regarded as a paragraph;

[0105] In this embodiment, the word count segmentation rule is to segment according to the number of words or terms in each paragraph or sentence. A word count threshold can be set based on experience. For example, every 500 words are segmented as a paragraph. This ensures that the length of each paragraph is relatively uniform, which is convenient for subsequent processing.

[0106] In this embodiment, the topic transition segmentation rule is to segment the text when the topic of the content changes significantly, such as from introducing background information to raising questions, or from explaining principles to giving examples. For example, in an article discussing environmental protection, when the author changes from introducing problems to proposing solutions, it can be segmented as a paragraph.

[0107] The working principle and beneficial effects of the above technical solution are: based on flexible segmentation rules, it can be adjusted and applied according to different text characteristics and learning needs, making the organizational structure of knowledge more flexible and closer to the actual needs of learners, thereby improving the efficiency and quality of training management learning.

[0108] Example 5

[0109] The present invention provides an online management method for training, which adjusts the size of the window to place the corresponding last matching segment, including:

[0110] Determine the sentence length of each text sentence in the last matching segment of each window, and at the same time, determine the initial size of each window;

[0111] The initial size is adjusted according to the sentence length to obtain the adjusted window size:

[0112] ;in, is the adjusted window size of the i-th window, is the average sentence length of all text sentences in the i-th window, is the length of the qth segment text to be placed in the i-th window, is the average segment length of all segmented texts required to be placed in the i-th window, n1 is the number of segments required to be placed in the i-th window, is the influence control factor of the i-th window on the q-th segment text, is the preset weight coefficient corresponding to the average sentence length in the i-th window, is the preset weight coefficient corresponding to the average segment length under the i-th window, is the initial size of the i-th window, is the average number of sentences involved in each segment text under the i-th window, For all The minimum length in .

[0113] In this embodiment, the initial size of the window refers to the initial size of each window before the window size is adjusted, which is determined based on parameters such as the average sentence length of all text sentences in each window, the length of the segmented text to be placed in each window, and the average segment length of all segmented text to be placed in each window.

[0114] In this embodiment, the influencing control factor refers to a factor that affects the size of each window after adjustment, and the size of the window is adjusted according to specific circumstances;

[0115] In this embodiment, the preset weight coefficient is used to adjust the weight of the sentence length and segment length on the window size during the window size adjustment process. It is set according to the actual situation to better adapt to different text contents and learning needs. For example, a window contains an article introducing a certain topic. If the concepts involved in the topic are more complex and require more explanations and examples, the impact control factor can be adjusted to a larger value so that more content can be placed in the window. At the same time, the preset weight coefficient can be adjusted to pay more attention to sentence length to ensure that the key content in the article can be fully explained and elaborated.

[0116] The working principle and beneficial effects of the above technical solution are: the window size can be flexibly adjusted according to the characteristics and requirements of different texts, taking into account the specific needs and influencing factors of each window, and can adapt to text contents of different types and lengths, thereby improving the adaptability and versatility of the management method, and by adjusting the window size, the available space can be maximized, and the efficiency of text display and learning can be improved. The efficiency of training management is improved.

[0117] Example 6

[0118] The present invention provides an online management method for training, which constructs a vector of a corresponding window based on all text sentences involved in each window, including:

[0119] Extract the vocabulary of the text sentences involved in each window;

[0120] Creating a vocabulary index for each vocabulary based on a preset vocabulary-index data table, and then building a vocabulary table corresponding to each sentence based on a preset table building method;

[0121] Calculate the TF-IDF value for each word:

[0122] ;in, is the TF-IDF value of the jth word in the dth text sentence in the ith window, is the preset word frequency control coefficient of the jth word in the dth text sentence in the ith window, and , is the frequency of occurrence of the jth word in the dth text sentence in the ith window, is the word with the highest frequency in the dth text sentence in the ith window The frequency of occurrence, is the number of text sentences in the ith window that contain the same word as the jth word in the dth text sentence, is the total number of text sentences in the i-th window, is the preset word weight of the jth word in the dth text sentence in the ith window, is the length of the dth text sentence in the ith window;

[0123] The TF-IDF value of each word in each window is filled into the vocabulary in turn to obtain the sentence vector of the corresponding text sentence, and then the vector of the corresponding window is obtained.

[0124] In this embodiment, the preset vocabulary-index data table is a predefined table or data structure used to map vocabulary to unique index values;

[0125] In this embodiment, the vocabulary index refers to the index value corresponding to each vocabulary in the preset vocabulary-index data table. By converting the vocabulary into an index, subsequent calculation and processing can be performed more efficiently. For example, for the preset vocabulary-index data table, the index corresponding to the vocabulary "apple" is 0, the index corresponding to the vocabulary "orange" is 1, and so on;

[0126] In this embodiment, the preset table construction method refers to a method or process for constructing a preset vocabulary-index data table, including extracting vocabulary from a large corpus and assigning indexes to them so that the vocabulary can be quickly accessed and processed in subsequent text processing;

[0127] In this embodiment, the preset word frequency control coefficient is a coefficient used to adjust the TF-IDF value of the vocabulary to take into account the importance or frequency of each vocabulary in the text. The coefficient can be set according to the actual situation to better reflect the weight of the vocabulary. For example, if a certain vocabulary is very common in the text of a specific field, but rarely appears in the text of other fields, then its TF-IDF value can be adjusted by the preset word frequency control coefficient to more accurately reflect its importance. It is a coefficient used to adjust the TF-IDF value of the vocabulary, calculated according to the preset word frequency and occurrence frequency of the vocabulary. This coefficient can be adjusted according to the characteristics and needs of the text to better reflect the importance of the vocabulary in the text.

[0128] The working principle and beneficial effects of the above technical solution are: by using TF-IDF values ​​to represent the importance of words in text sentences and converting text sentences into vector representations, the text information of each window can be quantified and counted, which helps to understand and analyze the text content involved in the training process, and provides a basis and support for subsequent learning path planning, knowledge graph construction, etc. At the same time, through vector representation, functions such as rapid retrieval and similarity matching of text information can also be realized, which improves the efficiency and effectiveness of training management.

[0129] Example 7

[0130] The present invention provides an online management method for training, which constructs a learning path for the trainee according to the first evaluation result and the knowledge graph, including:

[0131] Based on the first assessment result, determining the starting level and learning needs of the student and several key features associated with the knowledge graph;

[0132] Mapping each associated key feature into the state space to obtain the estimated learning state of the learner based on each knowledge unit structure;

[0133] Determining, based on the estimated learning state, an optional learning action for the user based on the corresponding knowledge unit structure;

[0134] Determine the action selection strategy based on the optional learning actions and the preset habit function;

[0135] According to the user action selection strategy, effective learning actions are selected from the optional learning actions of each knowledge unit structure, and based on the sequence of the knowledge unit structures in the knowledge graph, the learning path of the student is obtained.

[0136] In this embodiment, the starting level of the learner refers to the knowledge, skills and understanding level that the learner has before learning a specific field. The determination of the starting level is usually achieved by evaluating the existing knowledge and skill level of the learner. For example, if a learner has mastered basic mathematical knowledge, then his starting level may be high and he can directly enter a deeper level of mathematical learning.

[0137] In this embodiment, learning needs refer to the knowledge or skill gaps that students need to fill in a specific learning area in order to achieve specific learning goals. Learning needs can be determined based on the student's starting level and learning goals. If a student wants to improve his writing skills but has some deficiencies in grammar, then his learning needs may include strengthening the study and practice of grammar knowledge;

[0138] In this embodiment, the associated key features refer to features or factors that are closely related to the student's starting level and learning needs, which help determine the student's learning path and learning actions. For example, for a student's mathematics learning, the associated key features may include basic mathematics knowledge, mathematics learning goals, mathematics learning style, etc.

[0139] In this embodiment, the state space represents the current state of the learner, i.e., the starting level and learning needs of the learner, and the state space of the learning environment is a set describing the current state of the learner;

[0140] In this embodiment, the estimated learning state refers to the estimated state or progress of the student in the learning process based on the student's starting level, learning needs and related key features. The estimated learning state can help determine the learning action and learning path suitable for the student. If a student has a higher starting level in a certain knowledge field, but a lower starting level in another knowledge field, then his estimated learning state in different fields may be different.

[0141] In this embodiment, optional learning actions refer to actions or strategies that students can choose to take based on their current learning status during the learning process to improve their learning outcomes and achieve their learning goals. For example, optional learning actions may include reading textbooks, taking online courses, doing exercises, participating in discussions, etc.

[0142] In this embodiment, the preset habit function and action selection strategy refer to a function or strategy for determining learning actions that is preset according to the learning needs and characteristics of the trainees. These functions or strategies can be adjusted and optimized according to the actual situation of the trainees to improve the learning effect and learning experience. For example, the preset habit function can determine the priority and order of learning actions according to the learning habits and preferences of the trainees to better meet the learning needs of the trainees.

[0143] In this embodiment, the action selection strategy refers to the rule or scheme for selecting an action according to the current state. Depending on the selected algorithm, the strategy for defining the agent can be a deterministic strategy or a probabilistic strategy.

[0144] The working principle and beneficial effects of the above technical solution are: by combining the starting level and learning needs of the students, as well as the key features in the knowledge graph, a personalized learning path can be formulated for each student, improving learning efficiency and effectiveness, mapping the associated key features into the state space, and based on the estimated learning state of each knowledge unit structure, the student's current learning state can be more accurately estimated, based on optional learning actions and preset habit functions, determining the action selection strategy can help the system more effectively select learning actions suitable for students and improve the students' learning effects, according to the students' learning status and the sequence of knowledge unit structures in the knowledge graph, select effective learning actions and construct learning paths to ensure that students learn in a systematic order, and provide a good data foundation for training management by providing personalized learning paths and appropriate learning actions.

[0145] Example 8

[0146] The present invention provides an online management method for training, which maps each associated key feature into a state space, including:

[0147] Determine a value range of each associated key feature based on the feature type corresponding to each associated key feature;

[0148] Defining the dimensions of the student's state space based on each associated key feature and the corresponding value range, wherein each dimension corresponds to a key feature;

[0149] Combining the value of each associated key feature with the value of each other associated key feature based on a preset combination method to determine all possible state combinations;

[0150] The state space is constructed based on all possible state combinations.

[0151] In this embodiment, the associated key features are key features related to the starting level and learning needs of the learners, including: Initial knowledge level: the starting knowledge level of the learners in a specific field, which can be represented by an assessment score or a standardized knowledge level. Learning preferences: the learners' preferences for different learning contents, such as learning speed, learning style, and preferred learning resource types. Learning needs: the learners' learning needs in different knowledge fields or topics, such as strengthening certain concepts, filling knowledge gaps, etc.;

[0152] In this embodiment, the feature type refers to the nature or type of these features. For example, the starting level of the learner may belong to a specific level (e.g., beginner, intermediate, advanced), and the learning needs may belong to different categories (e.g., programming, mathematics, language learning);

[0153] In this embodiment, the value range is for each associated key feature, and its possible value range needs to be determined, for example, the initial knowledge level dimension: the value range is [0, 100];

[0154] In this embodiment, the dimension of the student's state space is defined based on each associated key feature and the corresponding value range. The state space can be a vector containing multiple dimensions, each dimension corresponding to a key feature, such as: initial knowledge level, learning preference vector, learning demand vector. For the initial knowledge level, the evaluation score is directly used as a dimension in the state space. For learning preferences, different preference types are converted into binary vectors and used as part of the state space. For learning needs, different needs are encoded and integrated into the state space with other features. For example, a learning system defines a state space, which contains two key features: the student's initial knowledge level and learning preference. The value range of the initial knowledge level is 0 to 100, and the learning preference may include "reading", "video" and "interaction". Based on these two features, we can define the state space as follows: Initial knowledge level dimension: the value range is [0, 100]. Learning preference dimension: the value range is ["reading", "video", "interaction"].

[0155] In this embodiment, the preset combination method is based on the value of each associated key feature and the preset combination method to determine all possible state combinations, and the value of each feature is combined to generate possible states of the students. For example, if the starting level of the students has three values: elementary, intermediate, and advanced, and the learning needs have three values: programming, mathematics, and language learning, the preset combination method can be the Cartesian product of the starting level and the learning needs, and all possible state combinations obtained include (elementary, programming), (elementary, mathematics), (elementary, language learning), (intermediate, programming), etc.

[0156] In this embodiment, constructing the state space based on all possible state combinations means that all possible state combinations are regarded as elements of the state space to constitute the complete state space of the trainee. For example, all possible state combinations obtained based on a preset combination method constitute the state space of the trainee. For example, the possible states of the trainee include (primary, programming), (primary, mathematics), (intermediate, programming), etc.

[0157] The working principle and beneficial effects of the above technical solution are: by mapping each associated key feature into the state space, the system can effectively describe the state and characteristics of the trainees, determine the feature type for each associated key feature, and determine the value range based on this, which helps the system to accurately capture the learning status and needs of the trainees. Based on the associated key features and their value ranges, the state space dimensions of the trainees are defined, so that the system can fully describe the status of the trainees. The preset combination method can combine the value of each associated key feature with the values ​​of other features to determine all possible state combinations, and construct the state space of the trainees based on all possible state combinations, so that the system can fully cover the possible states and situations of the trainees. It brings beneficial effects such as personalized learning, precise guidance, and improved learning experience, and improves the efficiency and quality of online training.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online management method for training, characterized in that: include: Step 1: Create a knowledge graph based on the preset learning content; Step 2: Conduct a first assessment of the student's starting level and learning needs, and construct a learning path for the student based on the first assessment result and the knowledge graph; Step 3: Determine several corresponding management threads based on each student's learning path, and monitor the corresponding student's learning activities in real time based on the management threads and preset monitoring tools; management threads refer to a series of management and guidance measures determined according to the student's learning path, including course setting, learning task assignment, learning progress tracking, and learning feedback; Step 4: Conduct a second assessment of the learning activities of the corresponding trainees, adjust the training strategy and conduct trainee training management based on the results of the second assessment; Build a knowledge graph based on preset learning content, including: Extract knowledge text from preset learning content based on natural language processing technology; Decompose the knowledge text into content to establish the knowledge unit structure; the knowledge unit structure refers to the relationship and organizational structure between the knowledge units obtained by content decomposition, which is used to express the hierarchical relationship and mutual correlation between different knowledge units; Build a knowledge graph based on the knowledge unit structure and preset layout; Decompose the knowledge text into content and establish the knowledge unit structure, including: Segment the knowledge text based on preset segmentation rules; The number of windows is determined according to the number of segments of the knowledge text, the segment length of each segment text, and the number of text sentences involved in the segment text; a window refers to an area where the knowledge text is divided; Divide the segmented text according to the number of windows to obtain the last matching segment of each window, and adjust the size of the window to place the corresponding last matching segment; the last matching segment is the paragraph divided and included in each window during the window division process; Based on all the text sentences involved in each window, a vector of the corresponding window is constructed; the vector of the window is a mathematical vector representing the text sentences in the window, which is used to characterize the content characteristics of the window; Decompose the vector of each window to extract the key features of each window; Determine the core theme of each window based on its key features; Treat each core topic as a knowledge unit and determine the nodes of the knowledge unit structure based on each knowledge unit; Conduct dependency analysis on each core topic to determine the dependencies between different topics; Determine the connection relationship between nodes based on the dependencies between different topics; Construct a knowledge unit structure based on all nodes and the connection relationships between nodes; The window is resized to place the corresponding last matching segment, including: Determine the sentence length of each text sentence in the last matching segment of each window, and at the same time, determine the initial size of each window; The initial size of the window is adjusted according to the length of the sentence to obtain the adjusted window size: ； Among them, W1 i is the adjusted window size of the i-th window, is the average sentence length of all text sentences in the i-th window, c i,q is the length of the qth segment text to be placed in the i-th window, c i,avg is the average segment length of all segmented texts required to be placed in the i-th window, n1 is the number of segments required to be placed in the i-th window, β i,q is the influence control factor of the i-th window on the q-th segment text, δ i is the preset weight coefficient corresponding to the average sentence length in the i-th window, μ i is the preset weight coefficient corresponding to the average segment length under the i-th window, W0 i is the initial size of the i-th window, is the average number of sentences involved in each segment text under the i-th window, c i,min For all c i,q The minimum length of is the set constant; s1 is the adjustment function.

2. The online management method for training according to claim 1, characterized in that: The preset segmentation rules include: paragraph segmentation, punctuation segmentation, keyword segmentation, word count segmentation and topic conversion segmentation.

3. The online management method for training according to claim 2, characterized in that: Based on all the text sentences involved in each window, construct the vector of the corresponding window, including: Extract the vocabulary of the text sentences involved in each window; Creating a vocabulary index for each vocabulary based on a preset vocabulary-index data table, and then constructing a vocabulary table corresponding to each sentence based on a preset table construction method; wherein the preset vocabulary-index data table is a predefined table or data structure for mapping vocabulary to a unique index value; the preset table construction method refers to a method or process for constructing a preset vocabulary-index data table, including extracting vocabulary from a large corpus and assigning indexes to it; Calculate the TF-IDF value for each word: ； in, is the TF-IDF value of the jth word in the dth text sentence in the ith window, α i,d,j is the preset word frequency control coefficient of the jth word in the dth text sentence in the ith window, and α i,d,j ∈(0,1), TF i,d,j is the frequency of occurrence of the jth word in the dth text sentence in the ith window, is the word with the highest frequency in the dth text sentence in the ith window The frequency of occurrence, DF i,d,j N is the number of text sentences in the i-th window that contain the same word as the j-th word in the d-th text sentence, i,d,j is the total number of text sentences in the i-th window, w i,d,j is the preset word weight of the jth word in the dth text sentence in the ith window, L i,d is the length of the dth text sentence in the ith window; The TF-IDF value of each word in each window is filled into the vocabulary table in turn to obtain the sentence vector of the corresponding text sentence, and then the vector of the corresponding window is obtained.

4. The online management method for training according to claim 1, characterized in that: Constructing a learning path for the student according to the first assessment result and the knowledge graph, including: Based on the first assessment result, determining the starting level and learning needs of the student and several key features associated with the knowledge graph; Mapping each associated key feature to the state space, obtaining the estimated learning state of the learner based on each knowledge unit structure; wherein the state space represents the current state of the learner, that is, the starting level and learning needs of the learner; the estimated learning state refers to estimating the state or progress of the learner in the learning process based on the starting level, learning needs and associated key features of the learner; Determine the optional learning actions of the learner based on the corresponding knowledge unit structure based on the estimated learning state; wherein the optional learning actions refer to the actions or strategies that the learner can choose to take based on the current learning state during the learning process; Based on the optional learning actions and the preset habit function, determine the student's action selection strategy; wherein the preset habit function is used to determine the priority and order of learning actions according to the student's learning habits and preferences; the action selection strategy refers to the rules or plans for selecting actions according to the student's current state; According to the student's action selection strategy, an effective learning action is selected from the optional learning actions of each knowledge unit structure, and based on the sequence of the knowledge unit structures in the knowledge graph, the student's learning path is obtained.

5. The online management method for training according to claim 4, characterized in that: Map each associated key feature into the state space, including: Determine a value range of each associated key feature based on the feature type corresponding to each associated key feature; Defining the dimensions of the student's state space based on each associated key feature and the corresponding value range, wherein each dimension corresponds to a key feature; Combining the value of each associated key feature with the value of each other associated key feature based on a preset combination method to determine all possible state combinations; The state space is constructed based on all possible state combinations.

Images