Micro course generation method based on OCR (Optical Character Recognition) and server

Abstract

The invention discloses a micro course generation method based on OCR recognition and a server, and relates to the field of image data processing, and the method comprises the steps: collecting maintenance record document image data, and carrying out the optical character recognition processing of the maintenance record document image data, and obtaining digital text data; performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, and integrating the terminology information, the fault code information and the operation step information into structured knowledge data; obtaining a knowledge level parameter and a post information parameter of the target user, and performing difficulty grading on the structured knowledge data according to the knowledge level parameter to obtain difficulty grading knowledge data; performing knowledge range screening in the difficulty grading knowledge data according to the post information parameters to obtain target knowledge data; and selecting teaching materials matched with the target knowledge data from the knowledge resource library, and generating a micro course resource bundle according to the teaching materials. By implementing the method, the updating efficiency of the training data can be improved.

Description

Technical Field

[0001] The present application relates to the field of image data processing, and in particular to a micro-course generation method and server based on OCR recognition. Background Art

[0002] With the rapid development of urban rail transit systems, equipment maintenance training plays an increasingly important role in ensuring operational safety and improving service quality. As a highly specialized field, rail transit equipment maintenance involves a large amount of complex technical knowledge and operating specifications, which places higher demands on the training of maintenance personnel. With the rapid development of information technology today, intelligent training methods are gradually becoming an important trend in the development of the industry.

[0003] In related technologies, unified training materials and fixed courses can be used. The training department usually organizes centralized lectures regularly or provides standardized electronic courseware for maintenance personnel to learn. In the process of collating training materials, trainers need to manually screen and organize historical maintenance records, and then write training content based on experience. This method requires a lot of human resources.

[0004] However, since the preparation and updating of training content requires a lot of manual work, the update cycle of training materials is long, the updating efficiency of training materials is low, and it is difficult to reflect the latest maintenance experience in a timely manner, resulting in maintenance personnel being unable to obtain the latest maintenance knowledge and skills in a timely manner, affecting the quality and efficiency of maintenance work. Summary of the invention

[0005] The present application provides a micro-course generation method and server based on OCR recognition, which are used to improve the updating efficiency of training materials.

[0006] In the first aspect, the present application provides a micro-course generation method based on OCR recognition, which is applied to a server, and the method includes: collecting maintenance record document image data, performing optical character recognition processing on the maintenance record document image data, and obtaining digital text data; performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, extracting professional terminology information, fault code information and operation step information, and integrating the professional terminology information, the fault code information and the operation step information into structured knowledge data; obtaining the knowledge level parameters and job information parameters of the target user, and performing difficulty classification on the structured knowledge data according to the knowledge level parameters to obtain difficulty classification knowledge data; performing knowledge range screening in the difficulty classification knowledge data according to the job information parameters to obtain target knowledge data; selecting teaching materials matching the target knowledge data from the knowledge resource library, and generating a micro-course resource package based on the teaching materials.

[0007] By adopting the above technical solution, the image data of maintenance record documents is collected, processed into digital text data through OCR, and then natural language processing is performed using the preset bidirectional encoder representation model to integrate professional terms, fault codes, and operating steps into structured knowledge data. Then, the target knowledge data is screened out according to the target user's knowledge level and job information parameters, and finally a micro-course resource package is generated. The collected data provides a basis for subsequent processing, OCR processing realizes document digitization, model processing improves information integration effects, parameter screening ensures personalized training content, and the generated micro-course resource package improves the efficiency of training material generation, meets actual needs, and improves training effects.

[0008] In combination with some embodiments of the first aspect, in some embodiments, the step of performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, extracting professional terminology information, fault code information and operation step information, and integrating the professional terminology information, the fault code information and the operation step information into structured knowledge data specifically includes: inputting the digital text data into a preset bidirectional encoder representation model to generate a text vector containing contextual semantic information; extracting professional terminology information, fault code information and operation step information from the text vector; establishing a first mapping relationship between the fault code information and the professional terminology information, and a second mapping relationship between the fault code information and the operation step information, and determining fault knowledge-related data based on the first mapping relationship and the second mapping relationship; classifying and arranging the fault knowledge-related data, and establishing a hierarchical structure including fault type, fault characteristics and solutions according to the frequency of fault occurrence, maintenance difficulty and equipment importance; and connecting the hierarchical structure in series according to the order of fault processing in the solution to form structured knowledge data.

[0009] By adopting the above technical solution, digital text data is input into the preset bidirectional encoder representation model to generate text vectors, extract key information such as professional terms, establish mapping relationships to determine fault knowledge related data, classify and organize, and establish a hierarchical structure based on fault characteristics, and finally connect them into structured knowledge data. The vectors generated by the model take into account contextual semantics to improve the accuracy of information extraction, the mapping relationship makes the knowledge closely related, and the hierarchical structure makes the knowledge logic clear, which helps to improve the systematicness and coherence of the knowledge in the training materials, thereby improving the quality and effect of training and helping trainees better master maintenance knowledge.

[0010] In combination with some embodiments of the first aspect, in some embodiments, the steps of obtaining the knowledge level parameter and job information parameter of the target user, grading the difficulty of the structured knowledge data according to the knowledge level parameter to obtain difficulty graded knowledge data, and screening the knowledge range in the difficulty graded knowledge data according to the job information parameter to obtain the target knowledge data specifically include: reading the working years data, professional qualification certificate data and historical training record data of the target user from the user profile database, calculating the knowledge level score according to the pre-designed scoring rules, and using the knowledge level score as the knowledge level parameter; reading the job code, job responsibilities list and skill requirement list of the target user from the user profile database, and generating job information parameters including job feature identifiers; grading the difficulty of the structured knowledge data according to the knowledge level parameter, dividing the structured knowledge data into elementary, intermediate and advanced levels, and obtaining difficulty graded knowledge data; calculating the matching degree between the job feature identifier in the job information parameter and the knowledge tag in the difficulty graded knowledge data, sorting the knowledge content according to the matching score, and selecting the knowledge content whose matching degree exceeds the preset matching threshold as the target knowledge data.

[0011] By adopting the above technical solution, the target user's years of work experience and other data are obtained from the user profile database to calculate the knowledge level score, and the job-related information is read to generate job information parameters. The structured knowledge data is graded in difficulty and the knowledge range is screened to obtain the target knowledge data. Multi-dimensional data is obtained to accurately evaluate the user's knowledge level. The generated parameters are used for difficulty classification to match the content to the user's ability, and the range is screened to ensure relevance to the job requirements. Accurate evaluation and personalized customization improve the pertinence and practicality of training, allowing users to efficiently obtain matching knowledge and improve work ability and efficiency.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after selecting teaching materials matching the target knowledge data from the knowledge resource library and generating a micro-course resource package based on the teaching materials, the method also includes: constructing a fault scenario simulation practice unit based on the fault code information and the operation step information; setting multiple checkpoints in the fault scenario simulation practice unit, each of which corresponds to a key operation in the operation step information; recording the operation behavior data of the target user at each checkpoint, and comparing the operation behavior data with the standard operation steps in real time; when an erroneous operation of the target user is detected, extracting professional terminology information and operation step information related to the standard operation steps from the target knowledge data to generate prompt content, and sending it to the target client.

[0013] By adopting the above technical solution, a fault scenario simulation exercise unit is constructed based on fault code information and operation step information, in which checkpoints are set corresponding to key operations, and the target user's operation behavior data is recorded and compared with the standard operation. When an erroneous operation is detected, relevant information is extracted from the target knowledge data to generate prompt content and send it to the client. The constructed simulation unit can reproduce the real maintenance scenario, the checkpoints ensure the standardization of operations, and the record comparison can detect errors in time. The extraction of information from the target knowledge data to generate prompt content ensures the accuracy and pertinence of the prompt.

[0014] In combination with some embodiments of the first aspect, in some embodiments, when an erroneous operation of the target user is detected, after the step of extracting professional terminology information and operation step information related to the standard operation steps from the target knowledge data to generate prompt content, the method also includes: obtaining the score of the target user in the fault scenario simulation practice unit; adjusting the difficulty level and knowledge point distribution of the micro-course resource package according to the score.

[0015] By adopting the above technical solution, we can obtain the target user's score in the fault scenario simulation exercise unit, and adjust the difficulty level and knowledge point distribution of the micro-course resource package based on the score. The score directly reflects the user's learning effect and operation proficiency. Adjusting the difficulty level based on this can make the course always adapt to the user's ability and avoid the influence of too high or too low difficulty on learning enthusiasm. The adjustment of knowledge point distribution can focus on the user's weak links, strengthen key knowledge, make the course content more reasonable, and thus optimize the user's learning path.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after selecting teaching materials matching the target knowledge data from the knowledge resource library and generating a micro-course resource package based on the teaching materials, the method also includes: performing statistical analysis on the learning behavior data of multiple target users using the same micro-course resource package, extracting the knowledge point access frequency, average learning time and repeated learning times in the learning behavior data; annotating the importance of professional terminology information and operation step information in the structured knowledge data according to the knowledge point access frequency, the average learning time and the repeated learning times to obtain annotation information; and adjusting the presentation order and display time of the teaching materials in the micro-course resource package based on the annotation information.

[0017] By adopting the above technical solution, the learning behavior data of multiple target users using the same micro-course resource package are statistically analyzed, and information such as the frequency of knowledge point visits is extracted. Based on this, the importance of professional terms and operation steps in the structured knowledge data is annotated, and the presentation order and display duration of teaching materials in the micro-course resource package are adjusted based on the annotations. Statistical analysis of multi-user learning behavior data can explore the actual value of knowledge, and importance annotation provides a basis for adjusting teaching materials. Reasonable adjustment of presentation order and display duration can highlight key knowledge and attract user attention.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of selecting teaching materials that match the target knowledge data from the knowledge resource library and generating a micro-course resource package based on the teaching materials, the method also includes: collecting problem feedback submitted by the target user during the learning process, and associating problem points that appear more than a preset frequency with the target knowledge content in the target knowledge data; and updating the content structure of the micro-course resource package based on the target knowledge content.

[0019] By adopting the above technical solutions, we can collect user feedback, screen high-frequency problem points and associate them with target knowledge content, and update the content structure of the micro-course resource package. Problem feedback reflects the learning difficulties of users, and the determination of high-frequency problem points focuses on key issues. The association with target knowledge content achieves precise optimization, making the course content more in line with user needs, strengthening weak links, and making knowledge explanation more in-depth and thorough. It improves the quality of training materials, makes users learn more smoothly, effectively improves learning effects, enhances the pertinence and practicality of training, and helps users better master knowledge.

[0020] In a second aspect, an embodiment of the present application provides a server, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the server to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a server, enables the server to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a server, enable the server to execute the method described in the first aspect and any possible implementation of the first aspect.

[0023] It is understandable that the server provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, which will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This application collects maintenance record document image data, processes it into digital text data through OCR, and then uses a preset bidirectional encoder representation model for natural language processing, integrating professional terms, fault codes, and operating steps into structured knowledge data, and then filters out target knowledge data based on the target user's knowledge level and job information parameters, and finally generates a micro-course resource package. The collected data provides a basis for subsequent processing, OCR processing realizes document digitization, model processing improves information integration effects, parameter screening ensures personalized training content, and the generated micro-course resource package improves the efficiency of training material generation, meets actual needs, and improves training effects.

[0025] 2. The present application constructs a fault scenario simulation exercise unit based on fault code information and operation step information, in which checkpoints are set corresponding to key operations, and the target user's operation behavior data is recorded and compared with the standard operation. When an erroneous operation is detected, relevant information is extracted from the target knowledge data to generate prompt content and send it to the client. The constructed simulation unit can reproduce the real maintenance scenario, the checkpoints ensure the standardization of operations, and the record comparison can detect errors in time. The extraction of information from the target knowledge data to generate prompt content ensures the accuracy and pertinence of the prompts.

[0026] 3. This application obtains the target user's score in the fault scenario simulation exercise unit and adjusts the difficulty level and knowledge point distribution of the micro-course resource package based on the score. The score directly reflects the user's learning effect and operation proficiency. Adjusting the difficulty level based on this can make the course always adapt to the user's ability and avoid the impact of too high or too low difficulty on learning enthusiasm. The adjustment of the knowledge point distribution can focus on the user's weak links, strengthen key knowledge, make the course content more reasonable, and thus optimize the user's learning path. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of a method for generating a micro-course based on OCR recognition in an embodiment of the present application; Figure 2 It is another flow chart of the method for generating a micro-course based on OCR recognition in an embodiment of the present application; Figure 3 It is another flow chart of the method for generating a micro-course based on OCR recognition in an embodiment of the present application; Figure 4 It is a schematic diagram of a physical device structure of a server in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0029] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0030] It should be noted that, currently, in urban transportation tracks, equipment maintenance is a complex and highly specialized task involving a large number of mechanical equipment of different types and specifications. Traditional maintenance training often relies on paper manuals or fixed electronic tutorials, which are slow to update and difficult to adapt to the rapidly changing equipment technology and the growing demand for personalized training. With the development of information technology, especially in the application of optical character recognition (OCR) technology and large machine learning models, it has become possible to automatically process and understand unstructured documents, providing technical support for the intelligent generation of personalized work order analysis and training courses.

[0031] For ease of understanding, the application scenarios of the embodiments of the present application are introduced below.

[0032] At present, there are two common methods for data processing of maintenance work orders: one is to manually input data into the system, which is time-consuming, labor-intensive and error-prone; the other is to use simple text matching tools to automatically extract key words, but this method is not accurate and cannot deeply understand and parse the content of the work order. In addition, there are rule-based expert systems for generating specific types of maintenance guides, but this requires a large number of rules to be defined in advance, which has high maintenance costs and poor flexibility.

[0033] Taking the training of a subway company as an example, the training staff regularly collects paper maintenance records, manually screens valuable information, and compiles training materials or produces electronic courseware based on experience. This method consumes a lot of manpower and time. For example, when updating the training materials for signal system maintenance, the training staff must check a large number of paper records one by one, sort out the key information, and then compile them into a unified teaching material. However, due to the cumbersome manual operation and the long data update cycle, the training materials are often not updated long after the new maintenance experience appears. The maintenance staff can only operate based on the old knowledge. When faced with new faults, they may not be able to handle them in a timely and accurate manner, affecting the normal operation of the subway.

[0034] Still taking the subway company as an example, after adopting the present application method, the server automatically collects maintenance record document image data, such as obtaining signal system maintenance record images through scanning equipment. After OCR processing into digital text data, the bidirectional encoder representation model is used to extract professional terms, fault codes, operation steps and other information, and integrate them into structured knowledge data. According to the knowledge level parameters of maintenance personnel (such as years of work, training experience, etc.) and job information parameters (such as job codes, skill requirements, etc.), suitable target knowledge data are screened out to generate micro-course resource packages. In this way, signal system maintenance training materials can be quickly updated, and maintenance personnel can obtain the latest knowledge in a timely manner. For example, for newly emerging signal fault handling methods, they can quickly learn and apply them to actual work, improve maintenance efficiency, and ensure subway operation safety.

[0035] For ease of understanding, the following describes the process of the method provided by this implementation in combination with the above scenario. Figure 1 , which is a flow chart of a micro-course generation method based on OCR recognition in an embodiment of the present application.

[0036] S101, collecting maintenance record document image data, performing optical character recognition processing on the maintenance record document image data, and obtaining digital text data.

[0037] Among them, maintenance record documents refer to written materials that record the equipment maintenance process, including maintenance time, maintenance objects, fault phenomena, treatment methods and other information, usually in the form of paper documents, photos or scans. Image data refers to digital image files formed after digital scanning or photographing of these documents. Common formats include JPG, PNG, etc. Optical character recognition (OCR) is a technology that converts text in an image into editable text, which is achieved through steps such as image preprocessing, character segmentation, feature extraction and pattern recognition. Digital text data refers to machine-readable text information obtained after OCR processing.

[0038] This step is performed when it is necessary to convert paper maintenance records into a processable digital form. Specifically, the server first collects images of maintenance record documents through scanners, cameras and other devices, or obtains existing document scans from the document management system. The collected images are then preprocessed, including image denoising, tilt correction, binarization and other operations to improve image quality. The OCR engine (such as Tesseract) is then called to perform character recognition on the processed image, and the text information in the image is converted into an editable text format, including text content, location information, etc. Finally, the recognition results are post-processed, such as typo correction, format normalization, etc., to obtain standardized digital text data.

[0039] In some embodiments, document digitization can be achieved through an OCR method based on deep learning: first, a pre-trained document image classification model is used to identify the document type, and corresponding preprocessing strategies are adopted according to different types; then a text detection model based on a convolutional neural network is used to locate the text area in the document and generate text box coordinates; finally, an OCR recognition model based on a recurrent neural network is used to recognize the text content line by line and output digital text. Optionally, a traditional rule-based OCR method can also be used: first, the image is grayscaled and binarized, and a projection analysis method is used for layout analysis and row and column segmentation; then a template matching algorithm is used to identify individual characters; finally, post-processing is performed through dictionary matching and context analysis to improve recognition accuracy. It is understandable that other OCR technologies can also be used to achieve document digitization, which is not limited here.

[0040] S102, performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, extracting professional terminology information, fault code information and operation step information, and integrating the professional terminology information, the fault code information and the operation step information into structured knowledge data.

[0041] Among them, the bidirectional encoder representation model is a neural network architecture that can simultaneously consider the forward and backward context information of the text sequence. Representative models include BERT, RoBERTa, etc. Professional terminology information refers to professional vocabulary in the maintenance field, such as equipment names, part names, etc. Fault code information is standardized fault numbers and descriptions. Operation step information contains specific maintenance operation processes and methods. Structured knowledge data is a data set that organizes the extracted information in a predefined format.

[0042] This step is performed after obtaining digital text data to extract valuable maintenance knowledge from the text. Specifically, the server inputs the digital text obtained by OCR into a pre-trained bidirectional encoder model (such as BERT). The model first segments the text into words and converts each word into a word vector. Then, through a multi-layer Transformer structure, the word vector sequence is bidirectionally encoded, considering the contextual relationship of each word in the entire sequence. Finally, the vector representation of the text is output, which contains the semantic information of the text. Then, various types of information are extracted from the vector through the information extraction module, and the information is integrated into a structured format.

[0043] In some embodiments, information extraction can be achieved through a rule-based approach: first, a professional terminology dictionary, a fault code library, and an operation step template are established; then, relevant information is identified from the text through regular expressions and keyword matching; finally, the extracted information is organized into a structured form according to a predefined pattern. Optionally, a deep learning method can also be used: first, a pre-trained named entity recognition model is used to identify professional terms; then, a BERT-based sequence labeling model is used to identify fault codes and operation steps; finally, the relationship between entities is obtained through a relationship extraction model to construct a knowledge graph. It is understandable that other natural language processing technologies can also be used to achieve information extraction and structuring, which are not limited here.

[0044] S103: Acquire knowledge level parameters and job information parameters of the target user, and perform difficulty grading on the structured knowledge data according to the knowledge level parameters to obtain difficulty graded knowledge data.

[0045] Among them, the target user refers to the individual maintenance personnel who need to be trained. The knowledge level parameter refers to the indicator used to quantitatively evaluate the user's professional knowledge mastery, including information such as years of work experience, certification status, and training experience. The position information parameter is used to represent the characteristics of the user's job position, including data such as job code, scope of responsibilities, and skill requirements. Structured knowledge data refers to a collection of maintenance knowledge content that has been systematically organized. Difficulty grading refers to the process of dividing knowledge content into different levels according to the degree of complexity. Difficulty graded knowledge data represents a knowledge system organized according to difficulty levels.

[0046] After obtaining structured knowledge data, this step is performed when personalized knowledge customization is required based on user characteristics. Specifically, the server first reads data such as the target user's years of work, professional qualification certificates, and historical training records from the user profile database. Then, according to the preset scoring rules, these data are converted into standardized knowledge level scores. At the same time, information such as the user's job code, job responsibilities list, and skill requirements are obtained to generate job characteristic data. Then, the knowledge level score is used to evaluate the difficulty of the structured knowledge data, and the knowledge content is divided into three levels: elementary, intermediate, and advanced, forming a knowledge system with graded difficulty.

[0047] In some embodiments, knowledge level assessment and difficulty grading can be achieved in a variety of ways: Optionally, a rule-based scoring method is used: first, the basic scores of years of work, certificate level, and training results are set; then the weighted total score is calculated according to the weights of each score; finally, the total score is mapped to the score ranges of different difficulty levels. Optionally, a machine learning method is used: first, a large number of users' knowledge level data and performance evaluation data are collected; then a classification model is trained to learn the correspondence between knowledge level and difficulty level; finally, the model is used to predict the knowledge level and match the content difficulty of new users. It is understandable that other evaluation methods can also be used to achieve knowledge difficulty grading, which is not limited here.

[0048] S104. Screen the knowledge range in the difficulty-graded knowledge data according to the job information parameters to obtain target knowledge data.

[0049] Among them, the job information parameter represents the characteristics of the user's job position, including job category, job responsibilities, skill requirements and other dimensions. Difficulty-graded knowledge data refers to the collection of knowledge content divided according to difficulty levels. Knowledge scope screening refers to the process of selecting relevant knowledge based on job requirements. Target knowledge data refers to the collection of knowledge content that matches the job after screening. Matching degree refers to the degree of relevance between knowledge content and job requirements.

[0050] After completing the knowledge difficulty classification, this step needs to be performed when the scope of training content needs to be determined based on job characteristics. Specifically, the server first converts the job information parameters into a standardized feature vector, including job category code, responsibility project code, skill level code, etc. Then, in the difficulty classification knowledge data, the similarity between each knowledge unit and the job feature vector is calculated to generate a matching score. Then, based on the preset matching threshold, the knowledge content with a high matching degree is screened out to ensure that the selected knowledge is highly relevant to the job requirements. Finally, the screened knowledge content is organized into a target knowledge data set.

[0051] In some embodiments, knowledge scope screening can be implemented in a variety of ways: Optionally, a rule-based mapping method is used: first, a position-knowledge correspondence table is established to define the knowledge items required for each position; then, the corresponding knowledge list is queried according to the user's position code; finally, cross-screening is performed in combination with the difficulty level to obtain the final knowledge set. Optionally, a semantic matching method is used: first, the position description and knowledge content are converted into semantic vectors; then, the cosine similarity between the vectors is calculated; finally, the matching knowledge content is screened according to the similarity threshold. It is understandable that other screening methods can also be used to determine the knowledge scope, which is not limited here.

[0052] S105: Select teaching materials that match the target knowledge data from the knowledge resource library, and generate a micro-course resource package based on the teaching materials.

[0053] Among them, the knowledge resource library refers to a data system that stores various teaching materials, including multimedia resources such as text, pictures, videos, animations, etc. Teaching materials refer to specific teaching content carriers, such as knowledge point explanation documents, operation demonstration videos, exercises, etc. Target knowledge data is used to represent a collection of screened knowledge content related to the position. Matching degree refers to the degree of relevance between teaching materials and target knowledge. The micro-course resource package represents a complete learning unit, including teaching objectives, learning content, practice tests, evaluation plans, etc. Teaching rules refer to the way of organizing teaching content that follows cognitive rules.

[0054] After determining the target knowledge data, this step is performed when a specific training course needs to be constructed. Specifically, the server first divides the target knowledge data into several learning units according to the knowledge points. For each learning unit, relevant teaching materials are retrieved from the knowledge resource library, including text explanations, graphic illustrations, video demonstrations, etc. The selected teaching materials must meet the requirements of complete knowledge point coverage, diverse expression forms, and moderate difficulty levels. The selected materials are organized according to the teaching rules to form a standard course structure consisting of five links: "introduction-explanation-example-exercise-summary". The duration of each link is controlled within 5 to 10 minutes to ensure the learning effect. Finally, all resources are packaged to generate a complete micro-course resource package containing course information, learning objectives, teaching content, and assessment standards.

[0055] In some embodiments, teaching material matching and micro-course generation can be achieved in a variety of ways: Optionally, a matching method based on vector similarity is adopted: first, the target knowledge content and teaching materials are converted into feature vectors; then the similarity scores between the vectors are calculated, and the materials with the highest similarity are selected; finally, the content is organized according to the teaching design template to generate a standardized course package. Optionally, an association method based on a knowledge graph is adopted: first, a domain knowledge graph is constructed to establish an association relationship between knowledge points and teaching materials; then, relevant materials are searched according to the position of the knowledge points in the graph; finally, the course structure is designed according to the learning path to form a systematic learning resource package. It is understandable that other methods can also be used to achieve teaching material selection and course generation, which are not limited here.

[0056] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , is another flow chart of the micro-course generation method based on OCR recognition in an embodiment of the present application.

[0057] S201, collecting maintenance record document image data, performing optical character recognition processing on the maintenance record document image data, and obtaining digital text data.

[0058] Maintenance record documents refer to written materials that record the equipment maintenance process, including maintenance time, maintenance objects, fault phenomena, and treatment methods. They are usually in the form of paper documents, photos, or scans. Image data refers to digital image files formed by digitally scanning or photographing these documents. Common formats include JPG, PNG, etc. Optical character recognition (OCR) is a technology that converts text in an image into editable text. It is achieved through steps such as image preprocessing, character segmentation, feature extraction, and pattern recognition. Digital text data refers to machine-readable text information obtained after OCR processing.

[0059] In the specific implementation process, the server first collects images of maintenance record documents through scanners, cameras and other devices, or obtains existing document scans from the document management system. Then the collected images are preprocessed, including image denoising, tilt correction, binarization and other operations to improve image quality. Then the OCR engine (such as Tesseract) is called to perform character recognition on the processed image, and the text information in the image is converted into an editable text format, including text content, location information, etc. Finally, the recognition results are post-processed, such as typo correction, format normalization, etc., to obtain standardized digital text data.

[0060] S202: Input the digital text data into a preset bidirectional encoder representation model to generate a text vector containing contextual semantic information.

[0061] The bidirectional encoder representation model is a neural network architecture that can simultaneously consider the forward and backward context information of the text sequence. Representative models include BERT, RoBERTa, etc. Contextual semantic information refers to the meaning of words in a specific context, including the dependency relationship and semantic association between words. The text vector is a numerical representation of the text, which can capture the semantic features of the text and facilitate subsequent processing.

[0062] The server inputs the digital text obtained by OCR into a pre-trained bidirectional encoder model (such as BERT). The model first segments the text into words and converts each word into a word vector. Then, through a multi-layer Transformer structure, the word vector sequence is bidirectionally encoded, considering the contextual relationship of each word in the entire sequence. Finally, the vector representation of the text is output. The vector contains the semantic information of the text and the dimension is usually 768 or 1024. This vector representation retains the semantic relationship in the text, which is conducive to subsequent information extraction.

[0063] S203: extracting professional terminology information, fault code information and operation step information from the text vector.

[0064] Professional term information refers to professional vocabulary in the field of maintenance, such as equipment names, part names, etc. Fault code information is the standardized fault number and description. Operation step information contains specific maintenance operation processes and methods. All of this information is contained in different dimensions of the text vector.

[0065] The server extracts information from the text vector in the following ways: using a pre-trained named entity recognition model to identify professional terms, including equipment names, part names, tool names, etc., using regular expressions or rule matching to identify fault codes in standard formats, and using a sequence labeling model to identify the temporal relationship of operating steps to extract the complete operating process. For each type of information, a corresponding standardized format is established to ensure the consistency and availability of the extraction results. The extracted information will be used for subsequent knowledge construction and course generation.

[0066] S204: Establish a first mapping relationship between the fault code information and the professional terminology information and a second mapping relationship between the fault code information and the operation step information, and determine fault knowledge association data based on the first mapping relationship and the second mapping relationship.

[0067] Fault code information refers to the standardized fault number and description, such as "E001-Main motor overload". Professional terminology information includes standardized terms such as equipment name, part name, professional concepts, such as "inverter" and "bearing". Operation step information is the specific operation guidance during the maintenance process, including operation actions, tool use, parameter settings, etc. Mapping relationship refers to the corresponding relationship between two types of information, which is established through association rules or semantic similarity. Fault knowledge association data refers to the knowledge network formed by mapping relationship, which contains complete association information of fault code, related terms and processing steps.

[0068] In the specific implementation, the server first establishes the first mapping relationship between the fault code and the professional terms: by analyzing the professional terms appearing in the fault code description, the association between the fault code and the terms such as related equipment, parts, parameters, etc. is established to form a data structure of {fault code: [related term 1, related term 2,]}. Then the second mapping relationship between the fault code and the operation step is established: each fault code is associated with the corresponding sequence of processing steps to form a data structure of {fault code: [step 1, step 2,]}. Finally, the two mapping relationships are integrated to form fault knowledge association data centered on the fault code, including related terms and processing steps. The data structure is {fault code: {related terms: [term 1, term 2,], operation steps: [step 1, step 2,]}}.

[0069] S205: Classify and organize the fault knowledge-related data, and establish a hierarchical structure including fault types, fault characteristics and solutions according to the fault frequency, maintenance difficulty and equipment importance.

[0070] Fault frequency refers to the number of times a specific fault occurs in historical records, which is used to reflect the commonness of the fault. Maintenance difficulty is a comprehensive assessment based on factors such as operation complexity, required skill level, and time consumption. The importance of equipment is determined by the role of the equipment in the system and the scope of fault impact. Fault type is a classification of the nature of the fault, such as mechanical fault, electrical fault, etc. Fault characteristics include characteristic descriptions such as the manifestation of the fault and influencing factors. The solution is a complete maintenance plan that includes processing steps, required tools, precautions, etc. The hierarchical structure is a tree structure that organizes related information according to logical relationships.

[0071] The server classifies and organizes the fault knowledge-related data: first, the number of occurrences of each fault is counted and the frequency is calculated; the maintenance difficulty score is assessed based on the number of operation steps, technical requirements and other factors; the equipment importance score is determined based on the importance of the equipment function and the scope of impact. Then, the faults are classified according to these three dimensions and a three-layer structure is constructed: the first layer is classified by fault type; the second layer contains the specific fault characteristics under each type; and the third layer is the corresponding solution. The data structure is {fault type: {fault characteristics: {frequency: value, difficulty: value, importance: value, solution: content}}}.

[0072] S206 , connecting the hierarchical structures in series according to the order of fault processing in the solution to form structured knowledge data.

[0073] The troubleshooting sequence in the solution refers to the temporal dependency of maintenance operations, including steps that must be performed in sequence and steps that can be performed in parallel. Hierarchical structure concatenation refers to reorganizing the scattered hierarchical nodes according to the order of operation to form a directed knowledge graph. Structured knowledge data is a standardized knowledge representation that contains elements such as entities, relationships, and attributes.

[0074] The server serializes the solution part in the hierarchical structure: analyzes the dependencies of the operation steps in each solution, identifies the sequence of steps that must be executed in order; adds preconditions and subsequent step identifiers to each step; and marks parallel operations as steps at the same level. Then the hierarchical structure is reorganized according to the order of operations to generate structured data containing a complete knowledge link. The data structure is {knowledge unit: {type: value, feature: value, precondition: [], subsequent: [], parallel: [], attribute: {frequency: value, difficulty: value, importance: value}, content: value}}. This structure not only retains the knowledge classification system, but also reflects the sequential relationship of operations.

[0075] S207, reading the target user's years of work data, professional qualification certificate data and historical training record data from the user profile database, calculating a knowledge level score according to pre-designed scoring rules, and using the knowledge level score as a knowledge level parameter.

[0076] The user profile database is a structured data storage system that stores basic information and career development information of users. The working years data records the working time of users in relevant positions in years. The professional qualification certificate data contains information such as the certificate name, grade, and acquisition time, such as "Senior Maintenance Engineer Certificate". The historical training record data contains information such as the training course name, completion time, and assessment results. The pre-designed scoring rules are quantitative standards for evaluating the user's knowledge level, including the weights and calculation methods of various indicators. The knowledge level score is a quantitative representation of the user's mastery of professional knowledge, usually in percentage.

[0077] The server reads the target user information from the user profile database through SQL query. First, the working years data is extracted, and the basic score is calculated based on 10 points per year. Then the professional qualification certificate data is analyzed, with 15 points for the primary certificate, 25 points for the intermediate certificate, and 40 points for the advanced certificate. Then the historical training records are counted, with 5 points for each completed relevant course and 3 points for excellent results. Finally, the scores of each item are weighted according to the weight: the weight of the working years score is 30%, the weight of the qualification certificate score is 40%, and the weight of the training record score is 30%, and the final knowledge level score is calculated. The score ranges from 0 to 100 points, reflecting the user's professional knowledge level and serving as an important basis for subsequent knowledge grading.

[0078] S208. Read the target user's job code, job responsibilities list and skill requirement list from the user profile database, and generate job information parameters including job feature identifiers.

[0079] Position codes are unique codes that identify different job positions, such as "MT001" for maintenance technicians. Job responsibilities lists the specific job content and requirements of a position. Skill requirements lists specify the professional skills and level requirements required for a position. Position feature identifiers are standardized descriptions of key job features, including feature values ​​for professional fields, scope of responsibilities, skill requirements, and other dimensions. Position information parameters are structured representations of job features.

[0080] The server extracts job-related information from the user profile database. First, it reads the job code and parses the job category, level and other information from the code. Then it parses the job responsibilities list, extracts key responsibilities and converts them into standardized responsibilities codes. Then it analyzes the skill requirements list and maps the skill items to the predefined skill classification system. Finally, it integrates all the information to generate a job feature identifier that includes professional field code, responsibility scope code and skill level code. This identifier fully describes the characteristics of the job and provides a benchmark for subsequent knowledge matching.

[0081] S209: grade the difficulty of the structured knowledge data according to the knowledge level parameter, and divide the structured knowledge data into elementary, intermediate and advanced levels to obtain knowledge data with graded difficulty.

[0082] Primary, intermediate and advanced are the three levels of knowledge difficulty, corresponding to basic knowledge, advanced knowledge and professional knowledge respectively. Structured knowledge data contains organized professional knowledge content and its attributes. Difficulty-graded knowledge data is a knowledge system organized by difficulty level. Knowledge level parameters are quantitative indicators of user knowledge level.

[0083] The server grades the difficulty of structured knowledge data according to the knowledge level parameters. First, set the difficulty grading standard: 0 to 60 points for elementary level, 61 to 85 points for intermediate level, and 86 to 100 points for advanced level. Then analyze the characteristics of each knowledge unit in the structured knowledge data, including the complexity of professional terms, the complexity of operation steps, and the requirements of pre-knowledge. Then evaluate the difficulty value of each knowledge unit based on these characteristics. Then assign the knowledge units to the corresponding difficulty levels according to the difficulty values. Finally, within each difficulty level, organize the knowledge structure according to the association between knowledge units to form a complete difficulty graded knowledge system.

[0084] S210, calculating the matching degree between the job feature identifier in the job information parameter and the knowledge label in the difficulty graded knowledge data, sorting the knowledge contents according to the matching degree scores, and selecting the knowledge contents whose matching degree exceeds a preset matching threshold as the target knowledge data.

[0085] The job feature identifier is a standardized description code for the job's professional field, scope of responsibilities, skill requirements and other dimensional features. The knowledge label is the attribute mark of each knowledge unit in the difficulty-graded knowledge data, which contains information such as knowledge type, application scenario, skill requirements, etc. The matching degree calculation is to calculate the similarity between two sets of identifiers through a vector space model. The matching score is a quantitative value of the similarity between identifiers, ranging from 0 to 1. The preset matching threshold is the standard score for screening knowledge content, usually set to 0.7. The target knowledge data is a collection of knowledge content after matching and screening.

[0086] The server performs the matching calculation between the job feature identifier and the knowledge label: first, the job feature identifier is converted into a feature vector, with each dimension corresponding to a feature attribute. Then the knowledge label is also converted into a vector form. The matching score is obtained by calculating the cosine similarity of the two vectors. The specific calculation method is: divide the dot product of the two vectors by the product of the module lengths of each vector. After calculating the matching degree for all knowledge contents, sort them from high to low according to the matching score. Select the knowledge content with a matching score greater than 0.7 to form the target knowledge data set. These knowledge contents are highly relevant to the job requirements and are suitable as training content.

[0087] S211. Select teaching materials that match the target knowledge data from the knowledge resource library, and generate a micro-course resource package based on the teaching materials.

[0088] The knowledge resource library is a data system that stores various teaching materials, including multimedia resources such as text, pictures, videos, animations, etc. Teaching materials are specific teaching content carriers, such as knowledge point explanation documents, operation demonstration videos, exercises, etc. The micro-course resource package is a complete learning unit, including teaching objectives, learning content, practice tests, etc. The target knowledge data is the screened knowledge content related to the position.

[0089] The server selects teaching materials from the knowledge resource library and generates micro courses: First, the target knowledge data is divided into several learning units according to knowledge points. For each learning unit, relevant teaching materials are retrieved from the knowledge resource library, including text explanations, graphic illustrations, video demonstrations, etc. The selected teaching materials must meet the requirements of complete knowledge point coverage, diverse expression forms, and moderate difficulty levels. The selected materials are organized according to the teaching rules to form a standard course structure with five links: "introduction-explanation-example-exercise-summary". The duration of each link is controlled within 5 to 10 minutes to ensure the learning effect. Finally, all resources are packaged to generate a complete micro course resource package containing course information, learning objectives, teaching content, and assessment standards.

[0090] The following is a more detailed description of the process of the method provided by this implementation. Figure 2 , is another flow chart of the micro-course generation method based on OCR recognition in an embodiment of the present application.

[0091] S301. Select teaching materials that match the target knowledge data from the knowledge resource library, and generate a micro-course resource package based on the teaching materials.

[0092] It can be understood that this step is similar to step S105 and will not be described in detail here.

[0093] S302: Construct a fault scenario simulation exercise unit based on the fault code information and the operation step information.

[0094] Fault code information is standardized data including fault number, fault description, fault type, etc. Operation step information is specific operation guidance during the maintenance process, including tool usage, parameter setting, operation sequence, etc. Fault scenario simulation exercise unit is a virtual training environment used to simulate maintenance operations under real fault conditions, including equipment status display, operation interface, feedback mechanism, etc.

[0095] The process of building a fault scenario simulation practice unit on the server first establishes a three-dimensional equipment model to fully present the appearance structure and internal parts of the equipment. Then the fault code information is converted into a specific equipment status performance, including abnormal indications, alarm information, parameter deviations, etc. Then the operation step information is converted into an interactive operation interface to realize functions such as tool selection, parameter adjustment, and component disassembly and assembly. On this basis, a physical engine is added to realize collision detection and dynamic response between parts. Finally, an audio-visual feedback mechanism is set up to warn of incorrect operations and confirm correct operations. The entire practice unit is implemented using Web3D technology, which supports multi-angle observation and free operation.

[0096] S303: Set a plurality of checkpoints in the fault scenario simulation exercise unit, each of which corresponds to a key operation in the operation step information.

[0097] Checkpoints are monitoring nodes set in the operation process to verify the correctness and completeness of the operation. Key operations are the steps in the operation process that play a decisive role in troubleshooting, such as core parameter adjustment, key component replacement, etc. The purpose of setting checkpoints is to ensure that users complete maintenance operations according to standard procedures to avoid missing important steps.

[0098] The specific implementation of setting checkpoints in the fault scenario simulation exercise unit of the server first analyzes the operation step information and identifies the key operation nodes. For each key operation node, set trigger conditions and verification rules. Trigger conditions include the correctness of tool selection, rationality of operation timing, accuracy of parameter settings, etc. Verification rules include operation action determination, parameter range check, state change confirmation, etc. Mark the checkpoint location with highlight prompts in the three-dimensional scene. Each checkpoint is equipped with an independent data acquisition module to record the user's operation data. Logical associations are established between checkpoints to form a complete operation verification link.

[0099] S304: Record the target user's operation behavior data at each checkpoint, and compare the operation behavior data with the standard operation steps in real time.

[0100] Operation behavior data is a record of all user behaviors during the operation process, including operation selection, operation time, operation sequence and other information. Standard operation procedures are pre-defined correct operation procedures, including specific requirements and judgment criteria for each step. Real-time comparison is to conduct instant comparison and analysis between the user's actual operation and the standard procedure.

[0101] The server records and compares user operation behaviors. First, the data collection module is deployed at each checkpoint to record all user operation information. The collected data includes operation timestamp, tool selection record, operation action sequence, parameter adjustment value, etc. The collected operation behavior data is structured according to the predetermined format. At the same time, the standard operation step data is loaded to establish an operation specification comparison table. Through the data comparison algorithm, the differences between the user operation and the standard steps in terms of timing, accuracy, completeness, etc. are analyzed. The comparison results are used to evaluate the correctness of the operation and serve as the basis for the next step of feedback and guidance.

[0102] S305: When an erroneous operation of the target user is detected, professional terminology information and operation step information related to the standard operation step are extracted from the target knowledge data to generate prompt content, and sent to the target client.

[0103] Wrong operation refers to the behavior that deviates from the standard operation steps, including wrong operation sequence, wrong tool selection, wrong parameter setting, etc. Professional terminology information is an explanation of the professional terms related to the operation, including term definitions, usage scenarios, precautions, etc. Operation step information contains specific operation instructions, such as tool usage, parameter adjustment range, operation precautions, etc. Prompt content is guidance information generated for wrong operation, including error reminders, correct demonstrations and improvement suggestions. The target client is the terminal device used by the user to practice operation.

[0104] The server detects an erroneous operation and first identifies the error type, compares the user's actual operation with the standard steps, and determines the link and specific manifestation of the error. Retrieve knowledge content related to the current operation steps from the target knowledge data, and extract corresponding professional terminology explanations and detailed operation instructions. Generate targeted prompt content based on the extracted information, which includes three parts: error description, professional knowledge supplement, and correct operation guidance. The generated prompt content is pushed to the target client in real time via WebSocket, and displayed in a pop-up window or highlighted form at an appropriate location on the interface. The display duration of the prompt content is set to 15 seconds, and it will automatically close after the user confirms.

[0105] S306: Obtain the score of the target user in the fault scenario simulation practice unit.

[0106] The score is a quantitative evaluation of the user's performance during the practice, including indicators such as operation accuracy, completion time, and number of errors. The scoring in the fault scenario simulation practice unit is based on the preset scoring rules, and each user's operation is scored and counted. The score data is used to evaluate the user's learning effect and operation proficiency.

[0107] The specific implementation process of the server obtaining the score situation first counts the user's operation data at each checkpoint, including operation correctness, operation time, number of retries, etc. Set different score weights for each checkpoint, and the weight of important operation steps is higher. Calculate the basic score according to the accuracy of the operation: full marks for correct operation, partial marks for incorrect operation but corrected after prompt, and no marks for multiple incorrect operations. Consider the operation time factor, add points if completed within the standard time, and deduct points if overtime. Count the number of incorrect operations and deduct the corresponding points for each error. Finally, summarize the scores of each item and generate a detailed scoring report containing the total score and the scores of each item.

[0108] S307. Adjust the difficulty level and knowledge point distribution of the micro-course resource package according to the score.

[0109] The difficulty level of the micro-course resource package is divided into three levels: elementary, intermediate, and advanced. Each level corresponds to knowledge content of different depths. Knowledge point distribution refers to the arrangement and proportion of knowledge content in the course. Difficulty adjustment refers to adjusting the depth and breadth of teaching content according to user performance. Knowledge point distribution adjustment refers to optimizing the arrangement order of knowledge content and the degree of detail of explanation.

[0110] The server adjusts the micro-course according to the score situation. First, analyze the distribution of users' scores on different knowledge points and identify the knowledge points that are not mastered enough. Determine the adjustment direction according to the score level: when the total score is lower than 60 points, reduce the difficulty level and supplement the basic knowledge explanation; when the total score is between 60 and 85 points, maintain the current difficulty and strengthen the weak knowledge points; when the total score is higher than 85 points, increase the difficulty level and add in-depth content. When adjusting the distribution of knowledge points, increase the length of explanation and the number of exercises for knowledge points with lower scores, and appropriately simplify the knowledge points that are well mastered. Reorganize the order of course content to ensure that the knowledge points are reasonably connected and the difficulty is steadily progressive. Generate a new micro-course resource package and update the course outline, learning objectives and assessment standards.

[0111] S308: Statistically analyze the learning behavior data of multiple target users using the same micro-course resource package, and extract the knowledge point access frequency, average learning time, and number of repeated learning times in the learning behavior data.

[0112] Learning behavior data is a collection of data that records various operations of users during the learning process, including access records, stay time, number of repetitions, and other information. Knowledge point access frequency refers to the number of times a specific knowledge content is viewed by users. Average learning time is the average learning time of users on each knowledge point. Repeated learning times are statistics of users' repeated learning behavior for the same knowledge point. Statistical analysis is the process of performing numerical calculations and extracting patterns from these data.

[0113] The server's statistical analysis of learning behavior data first extracts the learning records of all target users from the database. Count the number of visits to each knowledge point, record the visit timestamp and visit duration. Calculate the average daily visits and total visits to each knowledge point, and obtain the knowledge point visit frequency data. Analyze the user's stay time on each knowledge point, and calculate the average learning time after eliminating outliers. Identify users' repeated visits to knowledge points, count the number of times a single user has learned the same knowledge point, and calculate the number of repeated learning times. Generate a statistical report containing the visit frequency distribution, duration distribution, and repetition rate distribution.

[0114] S309: annotate the importance of the professional terminology information and the operation step information in the structured knowledge data according to the access frequency of the knowledge point, the average learning time and the number of repeated learning times to obtain annotation information.

[0115] Importance annotation is a quantitative identification of the value of knowledge content. Professional terminology information is technical terms and their explanations. Operation step information is the execution guidance of specific operations. Annotation information is a numerical representation of the importance of knowledge content, which is used to guide the subsequent content presentation optimization. Structured knowledge data is a collection of knowledge content that has been systematically organized.

[0116] The specific implementation of importance annotation by the server based on statistical data is to first establish an annotation scoring system and set the weight ratio of the three dimensions of access frequency, learning time, and number of repetitions. The access frequency data of each knowledge point is normalized and converted into a standard score of 0-100. The ratio of the average learning time to the preset standard time is calculated to generate a quantitative score for the time dimension. The number of repeated learning times is converted into a repetition rate indicator, and the repeated learning tendency score is calculated. Combining the scores of the three dimensions, the overall importance score of the knowledge point is calculated according to the weight ratio. The knowledge content is graded and annotated according to the score, and divided into three levels: core content, important content, and ordinary content. Generate an annotation dataset containing the knowledge point number, importance level, and score of each dimension.

[0117] S310: Adjust the presentation order and display duration of the teaching materials in the micro-course resource package based on the annotation information.

[0118] Teaching materials are specific carriers of knowledge content, including text, pictures, videos and other forms. Presentation order is the order in which teaching materials are arranged in the course. Display duration is the playback or display time of each teaching material. Annotation information is a quantitative indicator of the importance of knowledge content. Micro-course resource package is a complete learning course unit.

[0119] The specific implementation of the server adjusting the teaching materials based on the annotation information is to first read the importance level and score data in the annotation information. Prioritize the teaching materials according to the importance level, with the core content placed at the beginning of the unit, followed by the important content, and the ordinary content placed in the subsequent position. Adjust the display time of each teaching material, extend the display time of the core content, and appropriately compress the display time of the ordinary content. Rearrange the order of association between knowledge points to ensure that the previous and subsequent content are connected reasonably. Optimize the presentation of teaching materials, and add graphic illustrations and detailed explanations to important content. Generate a new material arrangement plan and update the content order and time schedule in the micro-course resource package.

[0120] S311, collecting the problem feedback submitted by the target user during the learning process, and associating the problem points that appear more than a preset frequency with the target knowledge content in the target knowledge data.

[0121] Problem feedback is the questions, suggestions and difficulty descriptions raised by users during the learning process, recorded in text form. The preset frequency is the quantitative standard for determining the importance of the problem, usually set to 10 times / month. Problem points are specific knowledge difficulties or operational difficulties extracted from problem feedback. Target knowledge content is the explanatory material in the knowledge base related to the problem point. The association relationship is the mapping connection between the problem point and the solution.

[0122] The specific implementation of the server collecting and processing problem feedback is to first collect the content of the questions submitted by users through the question-answering interface. Perform natural language processing on the collected question texts to extract keywords and core problem descriptions. Classify and aggregate the questions according to the topic, and count the submission frequency of each type of question. Filter out questions with a monthly submission frequency of more than 10 times and identify them as key questions. Retrieve knowledge content related to key questions from the target knowledge data, including concept explanations, operating instructions, precautions, etc. Establish a mapping table from problem points to knowledge content, and record information such as problem ID, knowledge point ID, and correlation. Store these associations in the database as a basis for subsequent content optimization.

[0123] S312. Update the content structure of the micro-course resource package according to the target knowledge content.

[0124] The content structure of the micro-course resource package is the organizational framework of the teaching content, including the division of knowledge units, the arrangement of content sequence, and the marking of key points and difficulties. The target knowledge content is the knowledge point material that needs to be strengthened. Content structure update refers to optimizing the organization of course content based on learning feedback.

[0125] The specific implementation of the server updating the micro-course content structure is to first load the original course structure data and identify the locations of the knowledge points that need to be strengthened. Add detailed concept explanations and example descriptions to the relevant knowledge points to expand the depth and breadth of the learning materials. Adjust the presentation of knowledge points and add charts and interactive examples to key content. Rearrange the order of knowledge units and explain the knowledge points corresponding to high-frequency questions in advance. Insert targeted exercises and self-test questions into the course to strengthen the consolidation of key knowledge. Update the course navigation and knowledge map to highlight the location marks of key content. Generate a new version of the course structure file and update the course outline and content index in the resource package. Record the updated version number and modified content for subsequent tracking and evaluation.

[0126] The following describes the server in the embodiment of the present invention from the perspective of hardware processing. Figure 4 , is a schematic diagram of a physical device structure of a server in an embodiment of the present application.

[0127] It should be noted that Figure 4 The structure of the server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0128] like Figure 4 As shown, the server includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403, such as executing the method described in the above embodiment. In RAM403, various programs and data required for system operation are also stored. CPU401, ROM402 and RAM403 are connected to each other through bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0129] The following components are connected to the I / O interface 405: an input section 406 including an audio input device, a button switch, etc.; an output section 407 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that a computer program read therefrom is installed into the storage section 408 as needed.

[0130] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 409, and / or installed from a removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the present invention are performed.

[0131] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0132] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0133] Specifically, the server of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the micro-course generation method based on OCR recognition provided in the above embodiment is implemented.

[0134] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the server described in the above embodiment; or may exist independently without being assembled into the server. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the server, the server implements the micro-course generation method based on OCR recognition provided in the above embodiment.

[0135] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0136] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0137] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. A micro-course generation method based on OCR recognition, characterized in that: Applied to a server, the method comprises: Collecting maintenance record document image data, performing optical character recognition processing on the maintenance record document image data to obtain digital text data; Performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, extracting professional terminology information, fault code information and operation step information, and integrating the professional terminology information, the fault code information and the operation step information into structured knowledge data; Acquire knowledge level parameters and job information parameters of the target user, and perform difficulty grading on the structured knowledge data according to the knowledge level parameters to obtain difficulty graded knowledge data; Screening the knowledge range in the difficulty graded knowledge data according to the job information parameters to obtain target knowledge data; Select teaching materials that match the target knowledge data from the knowledge resource library, and generate a micro-course resource package based on the teaching materials.

2. The method according to claim 1, characterized in that The step of performing natural language processing on the digital text data according to a preset bidirectional encoder representation model, extracting professional terminology information, fault code information and operation step information, and integrating the professional terminology information, the fault code information and the operation step information into structured knowledge data specifically includes: Inputting the digital text data into a preset bidirectional encoder representation model to generate a text vector containing contextual semantic information; Extracting professional terminology information, fault code information and operation step information from the text vector; Establishing a first mapping relationship between the fault code information and the professional terminology information, and a second mapping relationship between the fault code information and the operation step information, and determining fault knowledge association data based on the first mapping relationship and the second mapping relationship; Classify and organize the fault knowledge-related data, and establish a hierarchical structure including fault types, fault characteristics and solutions according to the frequency of faults, maintenance difficulty and equipment importance; The hierarchical structures are connected in series according to the order of fault processing in the solution to form structured knowledge data.

3. The method according to claim 1, characterized in that The steps of obtaining the knowledge level parameter and the job information parameter of the target user, grading the difficulty of the structured knowledge data according to the knowledge level parameter to obtain the difficulty graded knowledge data, and screening the knowledge range in the difficulty graded knowledge data according to the job information parameter to obtain the target knowledge data specifically include: Read the target user's working years data, professional qualification certificate data and historical training record data from the user archive database, calculate the knowledge level score according to the pre-designed scoring rules, and use the knowledge level score as the knowledge level parameter; Reading the target user's job code, job responsibilities list and skill requirement list from the user profile database, and generating job information parameters including job feature identifiers; According to the knowledge level parameter, the structured knowledge data is graded in difficulty, and the structured knowledge data is divided into elementary, intermediate and advanced levels to obtain knowledge data with graded difficulty; The matching degree between the position feature identifier in the position information parameter and the knowledge label in the difficulty graded knowledge data is calculated, the knowledge contents are sorted according to the matching degree score, and the knowledge contents whose matching degree exceeds a preset matching threshold are selected as the target knowledge data.

4. The method according to claim 1, characterized in that: After the step of selecting teaching materials matching the target knowledge data from the knowledge resource library and generating a micro-course resource package according to the teaching materials, the method further includes: Based on the fault code information and the operation step information, construct a fault scenario simulation exercise unit; Setting a plurality of checkpoints in the fault scenario simulation exercise unit, each of the checkpoints corresponding to a key operation in the operation step information; Recording the target user's operation behavior data at each of the checkpoints, and comparing the operation behavior data with the standard operation steps in real time; When an erroneous operation of the target user is detected, the professional terminology information and operation step information related to the standard operation steps are extracted from the target knowledge data to generate prompt content, and sent to the target client.

5. The method according to claim 4, characterized in that After the step of extracting professional terminology information and operation step information related to the standard operation steps from the target knowledge data to generate prompt content when an erroneous operation of the target user is detected, the method further includes: Obtaining the score of the target user in the fault scenario simulation practice unit; The difficulty level and knowledge point distribution of the micro-course resource package are adjusted according to the score.

6. The method according to claim 1, characterized in that After the step of selecting teaching materials matching the target knowledge data from the knowledge resource library and generating a micro-course resource package according to the teaching materials, the method further includes: Performing statistical analysis on the learning behavior data of multiple target users using the same micro-course resource package, and extracting the knowledge point access frequency, average learning time and number of repeated learning in the learning behavior data; According to the knowledge point access frequency, the average learning time and the number of repeated learning times, importance marking is performed on the professional terminology information and the operation step information in the structured knowledge data to obtain marking information; The presentation order and display duration of the teaching materials in the micro-course resource package are adjusted based on the annotation information.

7. The method according to claim 1, characterized in that After the step of selecting teaching materials matching the target knowledge data from the knowledge resource library and generating a micro-course resource package according to the teaching materials, the method further includes: Collecting the problem feedback submitted by the target user during the learning process, and associating the problem points that appear more than a preset frequency with the target knowledge content in the target knowledge data; The content structure of the micro-course resource package is updated according to the target knowledge content.

8. A server, characterized in that: The server includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the server to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a server, the server is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a server, the server is caused to execute the method according to any one of claims 1 to 7.

Images