Training course generation method and device

Abstract

The invention provides a training course generation method and device, and the method comprises the steps: obtaining a course generation text, inputting the course generation text to a course generation model, obtaining a to-be-trained course outputted by the course generation model, and enabling the to-be-trained course to be generated according to course structure description information, the course structure description information is generated according to the action structure information and a preset action library, and the action structure information is determined according to the course generation text. Due to the fact that the natural language is firstly analyzed into the structured middle layer, then the to-be-trained courses of different course types are uniformly constructed through the middle layer, the text is presented into the training elements which are usually dispersed in the character description paragraph, the structured middle layer is constructed and generated, and understanding, follow-up training and subsequent course generation of a user are facilitated.

Description

Technical Field

[0001] This application relates to the field of training course generation technology, and particularly to a method for generating training courses. This application also relates to a training course generation apparatus, a computing device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the development of mobile internet and social media, a large amount of fitness-related content is disseminated on various social media platforms in the form of summaries, notes, or captions. This content typically includes examples of training movements, exercise duration, rhythm arrangements, and training suggestions. Users often rely on this unstructured natural language text for understanding and following along during fitness training.

[0003] However, the above content is mostly presented in text, and the training elements are usually scattered in the text description paragraphs, which is not conducive to users' understanding, follow-up practice and subsequent course generation. Therefore, there is an urgent need for a course generation method to solve the above technical problems. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method for generating training courses. This application also relates to a training course generation apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address the aforementioned problems existing in the prior art.

[0005] According to a first aspect of the embodiments of this application, a training course generation method is provided, including: Obtain the course-generated text; The course-generated text is input into the course generation model to obtain the training course output by the course generation model. The training course is generated based on the course structure description information, which is generated based on the action structure information and the preset action library. The action structure information is determined based on the course-generated text.

[0006] According to a second aspect of the embodiments of this application, a training course generation apparatus is provided, comprising: The acquisition module is configured to acquire the course-generated text. The output module is configured to input the course-generated text into the course generation model to obtain the training course output by the course generation model. The training course is generated based on the course structure description information, which is generated based on the action structure information and a preset action library. The action structure information is determined based on the course-generated text.

[0007] According to a third aspect of the embodiments of this application, a computing device is provided, comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the above-described training course generation method.

[0008] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores a computer program / instructions, which, when executed by a processor, implement the steps of the training course generation method described above.

[0009] According to a fifth aspect of the embodiments of this application, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the training course generation method described above.

[0010] The training course generation method provided in this application obtains course generation text, inputs the course generation text into a course generation model, and obtains the training course output by the course generation model. The training course is generated based on course structure description information, which is generated based on action structure information and a preset action library. The action structure information is determined based on the course generation text.

[0011] The training course generation method provided in this application takes user-inputted natural language-based course generation text and feeds it into a course generation model. This text is then uniformly parsed into structured action structure information, mapped to existing action resources in a pre-defined action library, and used to generate course structure description information. Finally, the compliant course structure description information is used to generate structured training courses based on action types. Because the natural language is first parsed into action structure information at an intermediate layer, and then different course types are unified through this intermediate layer, the training course is constructed. Furthermore, by constructing the text training elements, which are typically scattered throughout the text description paragraphs, into a structured intermediate layer, it facilitates user understanding, practice, and subsequent course generation. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating a training course generation method provided in one embodiment of this application; Figure 2 This is a flowchart of a training course generation method provided in one embodiment of this application; Figure 3 This is a flowchart illustrating the action structure information of parsing course-generated text generation, provided in one embodiment of this application. Figure 4 This is a flowchart of generating action structure information provided in an embodiment of this application; Figure 5 This is a flowchart illustrating the generation of course structure description information based on a preset action library, provided in one embodiment of this application. Figure 6 This is a flowchart illustrating the verification process based on course compliance information, provided in one embodiment of this application. Figure 7 This is a flowchart of generating a training course provided in one embodiment of this application; Figure 8 This is a flowchart illustrating the generation and writing of a fused training course according to an embodiment of this application; Figure 9 This is a flowchart illustrating a method for generating training courses for shoulder training, provided in one embodiment of this application. Figure 10 This is a schematic diagram of the structure of a training course generation device provided in one embodiment of this application; Figure 11 This is a structural block diagram of a computing device provided in one embodiment of this application. Detailed Implementation

[0013] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0014] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0015] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0016] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0017] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0018] Large Language Models (LLMs), also known as large language models, can be understood as an advanced form of artificial intelligence based on deep learning. Trained on massive amounts of text data, they learn to understand and generate natural language similar to human language, predicting the next word and thus performing various tasks such as question answering, translation, summarization, and content creation. Large language models can have billions to trillions of parameters, representing the model's knowledge and capabilities. They can utilize multi-layered neural networks to simulate the human brain's information processing. Large language models are advanced artificial intelligence capable of understanding and creating human language, with wide applications and profoundly changing various industries.

[0019] Tabata training: Tabata is a highly efficient high-intensity interval training method. Its characteristics are that it involves cyclical exercise in a short period of time, which improves cardiopulmonary function and muscle strength, and burns fat efficiently.

[0020] With the development of social media and online fitness platforms, a large amount of fitness course content is published by fitness content creators in the form of natural language text (such as course descriptions, graphic notes, and subtitles). Users typically rely on this natural language text for course comprehension and follow-along training. However, due to the lack of structured data representation in this type of course text, while motion recognition and parameter extraction can be performed on the natural language text for conversion into standardized course data that can be accessed by exercise course systems, it is difficult to automatically construct a structured data model that meets the requirements of course execution. For example, it is difficult to automatically generate complete motion sequences, loop logic, rest points, and timeline nodes; manual organization and arrangement of the course are still required.

[0021] In actual course descriptions, circuit training and interval training structures such as Tabata often appear implicitly in natural language text. Existing automatic parsing methods struggle to accurately identify these hierarchical relationships and circuit structures, resulting in logically incomplete course outcomes that fail to accurately reflect the creator's course design intent. For running-related course texts, parameters such as pace zones, target heart rate, cadence, and lap count are often included. However, current technology struggles to automatically map these parameters to standardized training phases (such as warm-up, main training, and cool-down) and their corresponding time / distance configurations, hindering the standardized management and execution of running training courses.

[0022] Even with existing methods to structure the course text to some extent, the generated data often still lacks complete temporal information and parameter constraints, failing to meet the requirements of sports course platforms for course data structure. Consequently, it is difficult for the data to be directly integrated into the course library and used to generate online executable sports courses, reducing the efficiency of automated content production and reuse. Therefore, a training course generation method is urgently needed to solve the above-mentioned technical problems.

[0023] This application provides a method for generating training courses, and also relates to a training course generation apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0024] Figure 1 The diagram illustrates a flowchart of a training course generation method according to an embodiment of this application. After receiving course generation text input by the user, the course generation model parses the text to obtain action description information, further generates structured action structure information, matches it with existing action resources and action structure information in a preset action library to obtain standardized course structure description information, and then filters the course structure description information according to course compliance information to ensure the rationality and safety of the subsequently generated training courses, and to ensure that the course structure description information matches existing action resources. Finally, a corresponding training course is generated according to the course type (e.g., strength training course or running training course). The training course includes course meta-information such as course name, description, difficulty, and duration, as well as a course structure stored in the course platform for implementing an executable course. Different types of training courses can be further integrated.

[0025] Figure 2 The flowchart illustrates a training course generation method according to an embodiment of this application, which specifically includes the following steps: Step 202: Obtain the course-generated text.

[0026] Course-generated text can be understood as natural language text data used to describe the content of a sports course. Course-generated text can be text information input by the user, and can include text information that represents training arrangements or training plans, such as course titles, course introductions, text notes, and subtitles.

[0027] The large-scale model is pre-trained on a massive corpus and possesses strong natural language semantic understanding, context modeling, and structured information extraction and generation capabilities. Specifically, the large-scale model can semantically parse course texts containing colloquial expressions, non-standard units, missing subjects, or omitted steps, automatically inferring the relationships between implicit information such as training actions, duration, rhythm, and number of groups / repetitions. Based on this, it rewrites, completes, and reorganizes the original text to obtain target text that meets preset format requirements.

[0028] In one specific embodiment provided in this application, the user inputs the course generation text into the course generation model.

[0029] Taking shoulder training as an example, users can input the following text into the course generation model: "Shoulder training plan should include press, lateral raise, front raise, and rear deltoid training, such as seated barbell / dumbbell shoulder press (8-12 reps, 4-6 sets) to train the overall deltoid, dumbbell lateral raise (10-15 reps, 4-6 sets) to build shoulder width, dumbbell front raise (10-12 reps, 3-4 sets) to strengthen the anterior deltoid, and bent-over dumbbell / barbell reverse fly (or upright pull-up) to stimulate the rear deltoid. The movements should be slow, and the eccentric phase should be controlled to fully stimulate and shape a three-dimensional shoulder line."

[0030] The text acquisition solution based on large models can automatically adapt to the language styles and expression habits of different creators, and has better robustness and generalization ability to platform differences and differences in language, thus providing higher quality and more stable text input for subsequent action recognition, loop structure parsing and course modeling.

[0031] Step 204: Input the course-generated text into the course generation model to obtain the training course output by the course generation model. The training course is generated based on the course structure description information, which is generated based on the action structure information and the preset action library. The action structure information is determined based on the course-generated text.

[0032] The course generation model can be understood as an artificial intelligence model, used to generate structured courses for training. The courses for training can be understood as executable courses published on a course platform for user training, and can be in video format. Course structure description information can be understood as a unified and structured intermediate representation structure generated by a code generator; it can be a programmable data structure. Action structure information can be understood as structured action information generated based on user-input course generation text. A pre-set action library can be understood as a platform that pre-stores multiple action resources, which can be used to match the action structure information output by the course generation model with it to obtain a standardized course structure description.

[0033] The course generation model based on the large language model can automatically parse and process the text information input by the user, generate structured action structure information, and then convert the text information with greater randomness input by the user into course structure description information with unified expression standards through a preset action library. Finally, based on the course structure description information, a training course that is easy for the user to execute is generated.

[0034] In one specific embodiment of this application, after the course generation text is input into the course generation model, the course generation model outputs the corresponding training course. The training course is a structured video course, and the user can directly follow along with the training course.

[0035] Using the shoulder training example again, after receiving the user's input of shoulder training-related course generation text, the course generation model will generate a shoulder training course video. The video course contains multiple training stages, and each training stage includes multiple training movements. It can be seen that the video course is structured, and each training movement matches the movement resources in the preset movement library.

[0036] The course generation model can automatically output training courses. This method boasts a high degree of automation, replacing manual handwriting of action sequences, action loops, and action phases, significantly reducing labor costs. The course structure description information is obtained by mapping action structure information to a pre-defined action library. Therefore, all action steps are easy for users to understand and execute, and the rhythm of action execution can be estimated and controlled. The generated training courses can be used for user training, not just for demonstration. Furthermore, the generation of course structure description information is standardized, allowing for expansion to different types of sports courses without redesigning the underlying structure.

[0037] In one specific embodiment of this application, obtaining the training course output by the course generation model includes: The course generation text is parsed using the course generation model to obtain the action structure information; The course structure description information is generated based on the action structure information and the preset action library; A training course is generated based on the course structure description information and the action structure information.

[0038] If the course generation model directly outputs action text based on the course generation text, or simply breaks down actions, it cannot generate a structured training course. Therefore, it is possible to first generate structured action structure information as an intermediate layer based on the course generation text, and then map the action structure information to a preset action library to generate course structure description information. The action structure information can reflect the action type. Therefore, a structured and programmable training course can be generated based on the action structure information and the course structure description information.

[0039] In a specific embodiment provided in this application, the course generation text is parsed by a course generation model to obtain action structure information, course structure description information is generated based on the action structure information and a preset action library, and a course to be trained is generated based on the course structure description information and the action structure information.

[0040] Using the shoulder training example again, the course generation text is input into the course generation model to obtain the motion structure information for shoulder training. The motion structure information for shoulder training is then mapped to a preset motion library to obtain the course structure description information. Based on the course structure description information and the motion structure information, a training course for shoulder training is generated.

[0041] The course generation model can automatically output training courses. This method boasts a high degree of automation, replacing manual handwriting of action sequences, action loops, and action phases, significantly reducing labor costs. The course structure description information is obtained by mapping action structure information to a pre-set action library. Therefore, all action steps are easy for users to understand and execute, and the rhythm of action execution can be estimated and controlled. The generated training courses can be used for user training, not just for demonstration. Furthermore, the generation of course structure description information is consistent, allowing it to be extended to different types of sports courses without redesigning the underlying structure. Moreover, the method of generating training courses based on a structured intermediate layer improves the quality of the generated courses, solving the problem of poor course quality caused by the high degree of freedom and inconsistent expression in natural language.

[0042] In one specific embodiment of this application, the course generation text is parsed using the course generation model to obtain the action structure information, including: The course-generated text is parsed based on preset field standard information to obtain the action type and course initial text; The course reference text is obtained by parsing the action type. An information extraction template is determined based on the action type, and action description information is generated based on the information extraction template and the course reference text. The action structure information is generated based on the action type and the action description information.

[0043] Predefined field standard information can be understood as a set of predefined field configurations and specifications for various text information in the course-generated text. It can be used for the preprocessing and normalization of natural language text. Predefined field standard information can be used to remove irrelevant characters and noise content from the course text, and to segment and paragraph the text. Predefined field standard information can be used to uniformly convert units of different expressions into standard units used internally by the system, such as converting "minutes or seconds" to "seconds" and "meters or kilometers" to "kilometers". In addition, predefined field standard information can also be used to identify interjections such as "first...then...", "then", and "next", which indicate the sequence of actions or the connection between stages, and to convert colloquial expressions into serialized tags that can be recognized by the system, so that subsequent modules can stably build structured action descriptions.

[0044] Action type can be understood as the type of exercise described in the course-generated text. Action type can include running training type and strength training type. The course generation model can perform bidirectional classification of strength training and running training in the course-generated text. If running behavior, such as interval running, tempo running, 5-kilometer jogging, and other running training-related text information is identified, a dedicated parsing model for running training can be used to parse the course-generated text. Otherwise, a general parsing model can be used.

[0045] The initial course text can be understood as formatted text information generated using pre-defined standard field information. The course reference text can be understood as text information output by the parsing model corresponding to the action type. The information extraction template can be understood as a structured field template used to output structured description strings. It can include fields such as action name, metric type (e.g., number of repetitions, time, distance), capacity, number of sets, intervals, intensity, training phase, and cycle markers. The information extraction template can also be used to complete missing fields; for example, if the number of training sets is not specified, the information extraction template can automatically fill in the default number of sets. Action description information can be understood as structured descriptive data, which can serve as the formatted input data for the course generation model.

[0046] Because natural language text offers a high degree of freedom, different expressions can be standardized into a single internal representation by pre-setting standard field information, simplifying subsequent parsing logic. However, natural language text contains numerous sequence words and modal particles. If the natural language format is maintained, the course generation model will struggle to promptly and reliably identify the sequential relationships and loop structures between actions. Therefore, preprocessing and normalizing the natural language course text using pre-set standard field information yields a more regular and easily parsed initial course text.

[0047] To extract semantic information from the initial course text, a corresponding parsing model can be used based on the action type. This is because different types of actions have different characteristics, so using the corresponding parsing model can extract semantic features more accurately. Understandably, the course reference text obtained based on the parsing model has more significant semantic features.

[0048] Furthermore, structured field extraction and parameter completion can be achieved through information extraction templates, thereby transforming highly flexible natural language text into a structured set of fields. For missing fields, automatic completion can be performed based on the information extraction templates to ensure the structure and completeness of the output action structure description information.

[0049] For a detailed implementation of parsing course-generated text using a course generation model, please refer to [link to relevant documentation]. Figure 3 , Figure 3 This document illustrates a flowchart illustrating the process of generating action structure information from course-generated text according to an embodiment of this application. First, the course-generated text is parsed based on preset field standard information, removing colloquial expressions and converting fields to a uniform representation to obtain the initial course text and action types. Second, a corresponding parsing model is determined based on the action types to parse the initial course text, yielding course reference text. Finally, the action types can determine information extraction templates. Based on the information extraction templates and the course reference text, structured action description information can be generated, and then action structure information is generated based on the action types.

[0050] Continuing with the previous example, the generated course text is preprocessed and normalized based on preset field standard information to obtain the initial course text. Since the exercise type corresponding to the generated course text can be determined to be strength training, a general parsing model can be used to parse the initial course text. The course reference text can then be obtained through the parsing model. An information extraction template is then used to extract and complete the fields in the course reference text. Assuming that the generated course text does not include fields such as rest interval duration and exercise volume, information such as a 30-second rest interval and a 2kg dumbbell weight can be added based on the information extraction template. Finally, structured exercise description information is generated, followed by the generation of exercise structure information.

[0051] By converting the generated course text into a standardized course reference text using pre-defined standard information fields, the cost of designing and maintaining parsing rules for diverse natural language expressions is reduced. This also improves the robustness of subsequent action recognition, loop structure recovery, and course modeling, providing a reliable foundation for generating standardized course data that meets the platform's data structure requirements. The structured descriptions of the intermediate layer, constructed using information extraction templates, can be used to output programmable data structures. Furthermore, since strength training and running training have different characteristics, parsing models can be selected based on the action type, enabling automatic action type identification and classification. Using different parsing models generates structured action descriptions more suitable for different sports. The parsing results from different models can be automatically fused later, enabling the automatic generation of complex alternating training courses.

[0052] In a specific embodiment provided in this application, generating the action structure information based on the action type and the action description information includes: At least one training stage node is determined based on the action type and the action description information; Determine at least one action loop node corresponding to each training phase node, and determine at least one action node corresponding to each action loop node. The action structure information is generated based on each action node, each action loop node, and each training phase node.

[0053] Training phase nodes can be understood as sub-phases within a training course that have different purposes and focuses. For running-type training, training phase nodes may include warm-up, interval training, and cool-down phases. For strength training, training phase nodes may include warm-up, main training, and stretching phases.

[0054] Action loop nodes are used to chain together training actions, constructing action sequences into loop groups. It's understood that the training actions within a loop group can be the same or different. An action loop node can include at least one action node and a loop marker. The loop marker can represent the number of loops and the execution order of the training actions within the loop group. Action nodes can include parameters such as action name, capacity, and measurement method, corresponding to training phase nodes. If training phase nodes can be understood as describing the overall structure of the training course, then action nodes can be understood as its child nodes. Action structure information is a course logic tree generated sequentially from training phase nodes, action loop nodes, and action nodes. It can be a programmable data structure used to describe the execution order of the entire training course.

[0055] By converting natural language descriptions into a structured curriculum model containing multi-level nodes, training content presented in natural language can be abstracted into data objects that can be read, retrieved, and processed by computer programs, thereby supporting the automatic arrangement, verification, and reuse of curriculum content across multiple scenarios.

[0056] Through programmable data structures, on the one hand, the logical relationships implicit in natural language, such as the temporal relationships, cyclical relationships, and stage divisions of the course, can be expressed in explicit data form, which facilitates rule-based operations and logical reasoning by subsequent modules based on this structure; on the other hand, the course has a unified data interface with the platform's internal course engine, recommendation system, and statistical analysis module, and can be created, modified, combined, and scheduled as a unified data unit.

[0057] Building upon this foundation, the system can also split and reuse stages and sub-nodes across different courses, enabling the combination and replacement of course segments. Furthermore, it can adjust parameters or trim nodes based on user characteristics, available time, and training objectives, automatically generating course plans that meet personalized needs. Simultaneously, since the parameters of each course node are stored in structured fields, the system can also perform procedural checks on training intensity, duration of continuous high loads, and whether necessary warm-up or cool-down phases are included before course generation or execution, avoiding unreasonable training arrangements that are difficult to detect in a timely manner when relying solely on text-based script generation.

[0058] In one specific embodiment provided in this application, Figure 4 This is a flowchart of generating action structure information provided in an embodiment of this application. At least one training stage node is generated based on the action type and action description information, and a corresponding action loop node and at least one action node are generated for each training stage node.

[0059] Using the previous example, if the exercise type is strength training, then the exercise description information is divided into a warm-up phase, a formal training phase, and a stretching phase. In each training phase, taking the formal training phase as an example, the exercise cycle node reflects that the number of cycles is 3, that is, each exercise node is executed three times. The sequence of exercises is dumbbell shoulder press, dumbbell lateral raise, dumbbell front raise, and bent-over dumbbell reverse fly. The corresponding exercise nodes include 4, which include the name of each exercise, the volume of 2kg, and the number of repetitions. Based on the nodes of each training phase, the exercise cycle node, and each exercise node, the exercise structure information is generated. The exercise structure information can be a programmable data structure.

[0060] If the exercise type is running training, the exercise description information can be divided into training stages such as warm-up run, interval sprint, recovery run, and cool-down run. In each training stage, the exercise cycle node can be a single exercise or a cyclic exercise (alternating between sprint and jog). The exercise node can include structural information such as pace, heart rate zone, and cadence. Exercise structure information is generated based on the nodes of each training stage, the exercise cycle node, and each exercise node. Similarly, the exercise structure information is a programmable data structure.

[0061] By breaking down action description information into independently configurable training phase nodes, action loop nodes, and action nodes based on action type, the system can reuse existing training segments across different courses and supports the on-demand combination to generate new courses. When adjustments to a course are needed, only the corresponding phase or sub-node needs to be modified to complete the update, reducing course maintenance costs.

[0062] In one specific embodiment provided in this application, the action structure information includes at least one action node; The course structure description information is generated based on the action structure information and the preset action library, including: Determine the text features corresponding to each action node, and generate corresponding semantic vectors based on each text feature; Determine at least one candidate action corresponding to each semantic vector in the preset action library; Based on the candidate evaluation information, each candidate action is screened, and the target action and the target action identifier corresponding to the target action are determined. Update the corresponding action node based on the target action identifier, and generate the course structure description information.

[0063] See Figure 5 , Figure 5 This document illustrates a flowchart illustrating the generation of course structure description information based on a preset action library according to an embodiment of this application. Text features can be understood as a set of feature information representing each action node, reflecting the semantic information of each action node. Semantic vectors can be understood as multi-dimensional real-number vector representations that map text features to a large model or embedding model. Semantic vectors can be used to represent the semantic information of the text corresponding to action nodes in a continuous vector space, such as the action name, device, body part, and posture within the action node, as retrieval features. Correspondingly, semantically similar action nodes are closer in the vector space, facilitating subsequent similarity calculations, clustering, matching, or retrieval.

[0064] Candidate actions can be understood as keyword matching in a text retrieval engine and similarity retrieval in a vector retrieval library, and action resources obtained from the preset action library that match the semantic vector similarity of action nodes. Candidate evaluation information can be understood as evaluation information used to filter out suitable evaluation information for generating training courses, which may include evaluation elements such as comprehensive equipment consistency, posture rationality, part matching degree, and action style. Target actions can be understood as actions with better matching degree after filtering candidate actions based on a large language model. Target action identifiers can be understood as the meta-information of target actions. The corresponding course structure description information is a structured data description of semantic vector similarity matching of each action node in the preset action library.

[0065] Because natural language is highly flexible, different creators may use different expressions for the same action (for example, the same action could be described as "20 jumping jacks," "20 jumping jacks," or "20 jumping jacks"). Therefore, to generate playable training courses, action nodes need to be mapped to available action resources in a pre-set action library to obtain standardized target actions. Based on this, multiple dimensions of retrieval features, such as action name, device, body part, and posture, can be extracted from each action node, and corresponding semantic vectors can be generated for similarity matching. After selecting target actions from the pre-set action library based on candidate evaluation information, the target action identifier of the selected target action can be written into the action node for subsequent course generation. If there are no suitable resources in the pre-set action library, it will be marked as an unmatched node for subsequent course compliance information matching and coverage evaluation.

[0066] In a specific implementation of an embodiment of this application, the text features corresponding to each action node are determined, a corresponding semantic vector is generated based on each text feature, at least one candidate action corresponding to each semantic vector is determined in a preset action library, each candidate action is filtered according to the candidate evaluation information, a target action and the target action identifier corresponding to the target action are determined, the corresponding action node is updated according to the target action identifier, and course structure description information is generated.

[0067] Taking the bent-over dumbbell reverse fly as an example, in natural language, more professional terms include bent-over dumbbell rear deltoid fly and bent-over dumbbell rear deltoid lateral raise, while more colloquial terms include bent-over rear deltoid fly and bent-over fly for the rear deltoid muscles. If the action node is "bent-over rear deltoid fly," the corresponding text features are extracted, and a corresponding semantic vector is generated. In the preset action library, candidate actions include "bent-over dumbbell reverse fly," "bent-over dumbbell fly," and "bent-over lateral raise." Based on the candidate evaluation information, these three candidate actions are filtered to determine the target action as the bent-over dumbbell reverse fly. Then, its corresponding action identifier is obtained, and the corresponding action node is updated based on the action identifier, thereby generating course structure description information. By combining text retrieval, vector retrieval, and semantic comparison using a large language model, the most suitable action can be selected through name matching, action semantic matching, and then the large language model. Multi-dimensional matching based on the large language model can significantly improve the accuracy and coverage of action matching, thereby further ensuring the quality of the generated training courses.

[0068] In a specific embodiment provided in this application, a training course is generated based on the course structure description information and the action structure information, including: The course structure description information is verified based on the course compliance information to obtain the verification result; If the verification result is successful, a training course is generated based on the course structure description information and the action structure information.

[0069] Course compliance information can be understood as verification information that characterizes whether the course structure description information is suitable for generating a course to be trained. It is used to verify whether the course is safe, coherent, and executable before generating the course to be trained. The verification result can be understood as the result obtained after verification based on the course compliance information, which can include verification passed and verification failed.

[0070] Before generating a course to be trained, it is necessary to verify whether the course is safe, coherent, and executable. Therefore, the course structure description information can be verified through the course compliance information. If the verification passes, the course to be trained is generated. If the verification fails, the result is returned and the reason for failure is output.

[0071] In a specific implementation of one embodiment of this application, the course structure description information is verified according to the course compliance information to obtain a verification result. If the verification result is successful, a training course is generated based on the course structure description information and the action structure information.

[0072] Using the previous example, the course structure description information is judged to be safe, coherent and executable based on the course compliance information. The verification result is that the verification is passed. Therefore, the training course is generated based on the course structure description information and the action structure information.

[0073] Course compliance information can ensure improved course quality in the generated training courses.

[0074] In one specific embodiment provided in this application, the course compliance information includes security verification information, pacing verification information, and matching coverage information; The course structure description information is verified based on the course compliance information to obtain the verification results, including: The course structure description information is verified based on the security verification information to obtain a first verification result; If the first verification result is qualified, the course structure description information is verified according to the rhythm verification information to obtain the second verification result; If the second verification result is satisfactory, the course structure description information is verified according to the matching coverage information to obtain the verification result.

[0075] See Figure 6 , Figure 6 A flowchart illustrating a course compliance information verification process according to an embodiment of this application is shown. Safety verification information can be understood as verification information assessing the safety of action combinations and intensity, used to check for high-risk actions, high-intensity combinations, etc. Correspondingly, the first verification result can be understood as a course structure description information indicating that the action combinations and intensity meet safety standards. Rhythm verification information can be understood as verification information regarding the rationality of the course rhythm and training phases, used to determine if there are missing training phases (such as missing warm-up phases) and whether the action content matches the training phase (high-intensity actions appearing in the warm-up phase). Correspondingly, the second verification result can be understood as a course structure description information indicating that the actions are safe and the training rhythm is reasonable. Matching coverage information can be understood as verification information characterizing whether each action node matches actions in a preset action library. If the matching rate is lower than a preset matching threshold, it is determined that the course cannot be generated, and the user needs to readjust the input.

[0076] From the perspective of sports safety, the course generation model constrains and verifies the courses to be generated from multiple dimensions to ensure the safety and controllability of the generated sports courses during execution.

[0077] In a specific embodiment provided in this application, the course structure description information is verified according to the security verification information to obtain a first verification result. If the first verification result is qualified, the course structure description information is verified according to the rhythm verification information to obtain a second verification result. If the second verification result is qualified, the course structure description information is verified according to the matching coverage information to obtain a verification result.

[0078] Taking the generation of a shoulder training course as an example, the first step is to determine whether there are high-risk movements, dangerous combinations, or excessive loads in the course structure description information for the shoulder. If the first verification result is qualified, then it is determined whether the training phase is complete and whether the movement nodes in each training phase are consistent with the training phase to which they belong. If the second verification result is qualified, then it is determined whether the matching coverage rate of each movement node in the preset movement library is greater than the preset matching coverage rate threshold. If so, the verification result is qualified.

[0079] By automating closed-loop checks on course structure description information through multi-dimensional course compliance information, and employing multiple mechanisms such as action safety detection, intensity reasonableness detection, and action matching coverage, the efficiency of course generation is significantly improved, and the risk of generating non-compliant courses is significantly reduced.

[0080] In a specific embodiment provided in this application, a training course is generated based on the course structure description information and the action structure information, including: Generate course metadata based on the action structure information; Generate a course structure based on the course structure description information; The course to be trained is generated based on the course metadata and the course structure.

[0081] See Figure 7 , Figure 7 A flowchart illustrating the generation of a training course according to an embodiment of this application is shown. Course metadata can be understood as a set of basic information describing the overall attributes of the training course and presented intuitively to users during browsing and course selection. Accordingly, course metadata may include the course title, course cover image, course description text, total course duration, target audience (e.g., beginner / intermediate / runner), training goals (e.g., fat loss, muscle gain, body shaping), main training areas (e.g., upper limbs, lower limbs, core), required equipment type, course difficulty level, course style tags (e.g., HIIT, Tabata, strength training, etc.), and other information related to course display and retrieval. Course metadata typically does not directly determine the specific action execution process but is used for course categorization, display, and retrieval.

[0082] A course structure can be understood as a structured data set that describes the specific training content and execution logic of a course to be trained. Accordingly, a course structure can include multiple training phases (such as warm-up, main training, and cool-down phases), as well as action nodes, rest nodes, and cycle nodes within each phase. For each node, the course structure can record parameters such as action name, number of sets / repetitions, duration, pace zone, target heart rate zone, cadence, and rest duration, as well as the sequential relationships, cycle relationships, and timeline positions between nodes. The course structure is used to drive the actual execution and interactive control of the course on the terminal device.

[0083] By dividing the courses to be trained into two parts—course metadata and course structure—for generation and storage, the programmability, manageability, and reusability of the courses within the system can be improved while ensuring the user experience.

[0084] In a specific embodiment provided in this application, course meta-information is generated based on action structure information, course structure is generated based on course structure description information, and a course to be trained is generated based on the course meta-information and the course structure.

[0085] See Figure 8 , Figure 8 The diagram illustrates a flowchart of generating and writing a fusion training course according to an embodiment of this application. The course generation model receives course-generated text corresponding to strength training and running training; therefore, the input course-generated text can be from the same natural language or different natural languages. The courses are then combined and arranged in sequence according to the training objective; for example, if the training objective is fat loss, strength training can be performed first, followed by running training. The course metadata and structure are unified, and the corresponding training course is generated. The course metadata can include an overall course identifier and identifiers for each sub-course, used for front-end video playback and training statistics. The training course structure is complete and can be presented as a composite course to various users and written to the platform.

[0086] Taking the generation of a shoulder training course as an example, the course generation model can be used to generate a course name, description, difficulty, and course duration from the motion structure information. A course cover can be selected from the platform's image service based on the above information to construct course meta-information. Then, a course structure is generated based on the course description information to describe the specific training content and execution logic of the training course, thus obtaining the training course as shoulder training and writing it into the course library.

[0087] The course structure description information is mapped to the platform's course data structure to obtain the course structure. Structured data such as action step sequences, loop structures, and stage segments are generated. Then, a large model is invoked to generate course metadata based on the course structure information, including the course name and description, and to estimate the course difficulty and duration. The difficulty level is estimated based on the number of movements, training volume, and intensity parameters. The total training time is calculated based on duration and intervals, and a course cover image is generated. A service program for generating images is invoked to generate a course cover image based on the course type, training goals, and style. Finally, the course metadata and course structure are written into the platform's course library.

[0088] The generated training courses are represented by dividing them into two parts: course metadata and course structure. The course metadata primarily serves course display and selection scenarios, while the course structure primarily serves course execution and playback control. This separation allows for adjustments to the cover, description, tags, and other displayed content without altering the course structure. Furthermore, it enables optimization of the internal course action choreography while keeping the course metadata unchanged, facilitating course version maintenance and operational adjustments.

[0089] Course metadata describes macro-level attributes such as course objectives, target audience, duration, and difficulty in a tagged and structured format. This facilitates rapid filtering, sorting, and accurate recommendations by retrieval and recommendation systems based on this metadata. Simultaneously, the platform can perform statistical analysis and configure operational strategies based on this metadata without needing to parse the specific action-level structure. Course structure focuses on the timing and parameter configuration of training actions. Based on this structured course structure data, the system can automatically perform intensity verification, safety checks, and personalized trimming, replacement, and combination of courses. This structured representation, oriented towards execution logic, allows for the generation of personalized execution plans for different users while maintaining consistency in course metadata.

[0090] With the separation of metadata and structure, the same course structure can be reused in different display formats and terminal devices (such as mobile apps, tablets, TVs, or wearable devices). The front end only needs to render the corresponding timeline and action guides according to the course structure and complete the interface display in combination with the course metadata, thereby improving the system's scalability and reusability.

[0091] Multiple training courses can also be unified in terms of time and course logic to obtain integrated courses. For example, running courses and strength training courses can be integrated, and a single course data structure or a combination of multiple training courses that conform to the platform specifications can be output, thereby realizing the automated integration and unified distribution of multiple courses.

[0092] The training course generation method provided in this application takes user-inputted natural language-based course generation text and feeds it into a course generation model. This text is then uniformly parsed into structured action structure information, mapped to existing action resources in a pre-defined action library, and used to generate course structure description information. Finally, the compliant course structure description information is used to generate structured training courses based on action types. Because the natural language is first parsed into action structure information in an intermediate layer, and then different course types are unified through this intermediate layer, the training course is constructed. Furthermore, by constructing the text training elements, which are typically scattered throughout the text description paragraphs, into a structured intermediate layer, it facilitates user understanding, practice, and subsequent course generation.

[0093] The training course generation method provided in this application achieves low-cost, large-scale production of an unlimited amount of training course content through automatic course generation and structured modeling. This enables the construction of an AI course library and supports a personalized recommendation system, allowing users to generate courses independently and significantly enhancing interactivity and user retention. Furthermore, the course structure description information provided in this application can be directly used in course follow-up and training feedback systems, giving courses standardized characteristics such as executableness, traceability, and analyzability, and providing a key technological foundation for subsequent intelligent directions such as "intelligent training plans" and "AI personal trainers." Compared to the template-based generation of strength training or running, the training course generation method provided in this application achieves automatic stage identification for running courses and structured expression of key training parameters such as pace, heart rate, and cadence. It also achieves automatic fusion generation of running and strength training, thereby truly covering the user's full-link training needs from running to strength training, improving the adaptability, richness, and training guidance effect of course generation.

[0094] The following is in conjunction with the appendix Figure 9 Taking the application of the training course generation method provided in this application in shoulder training as an example, the training course generation method will be further explained. Figure 9 This application provides a flowchart illustrating a method for generating training courses for shoulder training, which includes the following steps: Step 902: The course generation model parses the shoulder training course generation text according to the preset field standard information to obtain the movement type as strength training and the course initial text.

[0095] Step 904: Determine the analytical model corresponding to strength training, parse the initial course text, and obtain the course reference text.

[0096] Step 906: Determine the information extraction template corresponding to strength training, and generate movement description information based on the course reference text and the information extraction template.

[0097] Step 908: Determine the three training phase nodes based on the type of strength training movement and movement description information.

[0098] Step 910: Determine at least one action loop node corresponding to each training phase, and at least one action node corresponding to each action loop node.

[0099] Step 912: Generate the action structure information based on each action node, each action loop node, and each training stage node.

[0100] Step 914: Determine the text features corresponding to each action node, and generate the corresponding semantic vector based on each text feature.

[0101] Step 916: Determine at least one candidate action corresponding to each semantic vector in the preset action library, filter each candidate action according to the candidate evaluation information, and determine the target action and the target action identifier corresponding to the target action.

[0102] Step 918: Update the corresponding action node according to the target action identifier to generate the course structure description information.

[0103] Step 920: Verify the course structure description information based on the course compliance information and obtain the verification result.

[0104] Step 922: Generate course meta-information based on the action structure information, generate course structure based on the course structure description information, and generate a course to be trained based on the course meta-information and the course structure.

[0105] The training course generation method provided in this application takes user-inputted natural language-based course generation text and feeds it into a course generation model. This text is then uniformly parsed into structured action structure information, mapped to existing action resources in a pre-defined action library, and used to generate course structure description information. Finally, the compliant course structure description information is used to generate structured training courses based on action types. Because the natural language is first parsed into action structure information in an intermediate layer, and then different course types are unified through this intermediate layer, the training course is constructed. Furthermore, by constructing the text training elements, which are typically scattered throughout the text description paragraphs, into a structured intermediate layer, it facilitates user understanding, practice, and subsequent course generation.

[0106] Corresponding to the above method embodiments, this application also provides embodiments of a training course generation apparatus. Figure 10 A schematic diagram of a training course generation device according to an embodiment of this application is shown. Figure 10 As shown, the device includes: Module 1002 is configured to retrieve course-generated text. The output module 1004 is configured to input the course-generated text into the course generation model to obtain the training course output by the course generation model, wherein the training course is generated based on the course structure description information, the course structure description information is generated based on the action structure information and the preset action library, and the action structure information is determined based on the course-generated text.

[0107] Optionally, the output module 1004 is further configured to: The course generation text is parsed using the course generation model to obtain the action structure information; The course structure description information is generated based on the action structure information and the preset action library; A training course is generated based on the course structure description information and the action structure information.

[0108] Optionally, the output module 1004 is further configured to: The course-generated text is parsed based on preset field standard information to obtain the action type and course initial text; The course reference text is obtained by parsing the action type. An information extraction template is determined based on the action type, and action description information is generated based on the information extraction template and the course reference text. The action structure information is generated based on the action type and the action description information.

[0109] Optionally, the output module 1004 is further configured to: At least one training stage node is determined based on the action type and the action description information; Determine at least one action loop node corresponding to each training phase node, and determine at least one action node corresponding to each action loop node. The action structure information is generated based on each action node, each action loop node, and each training phase node.

[0110] Optionally, the action structure information includes at least one action node; The output module 1004 is further configured as follows: Determine the text features corresponding to each action node, and generate corresponding semantic vectors based on each text feature; Determine at least one candidate action corresponding to each semantic vector in the preset action library; Based on the candidate evaluation information, each candidate action is screened, and the target action and the target action identifier corresponding to the target action are determined. Update the corresponding action node based on the target action identifier, and generate the course structure description information.

[0111] Optionally, the output module 1004 is further configured to: The course structure description information is verified based on the course compliance information to obtain the verification result; If the verification result is successful, a training course is generated based on the course structure description information and the action structure information.

[0112] Optionally, the course compliance information includes security verification information, pacing verification information, and matching coverage information; The output module 1004 is further configured as follows: The course structure description information is verified based on the course compliance information to obtain the verification results, including: The course structure description information is verified based on the security verification information to obtain a first verification result; If the first verification result is qualified, the course structure description information is verified according to the rhythm verification information to obtain the second verification result; If the second verification result is satisfactory, the course structure description information is verified according to the matching coverage information to obtain the verification result.

[0113] Optionally, the output module 1004 is further configured to: Generate course metadata based on the action structure information; Generate a course structure based on the course structure description information; The course to be trained is generated based on the course metadata and the course structure.

[0114] The training course generation device provided in this application obtains course generation text, inputs the course generation text into a course generation model, and obtains a training course output by the course generation model. The training course is generated based on course structure description information, which is generated based on action structure information and a preset action library. The action structure information is determined based on the course generation text.

[0115] The training course generation device provided in this application takes user-inputted natural language-based course generation text and inputs it into a course generation model. This text is then uniformly parsed into structured action structure information. Existing action resources are then matched with this action structure information to generate course structure description information. Finally, the compliant course structure description information is used to generate structured training courses based on action types. Because the natural language is first parsed into a structured intermediate layer, and then different types of training courses are uniformly constructed through this intermediate layer, the training elements, which are usually scattered throughout the text description paragraphs, are presented in a structured intermediate layer, facilitating user understanding, practice, and subsequent course generation. The above is an illustrative scheme of a training course generation device according to this embodiment. It should be noted that the technical solution of this training course generation device and the technical solution of the above-described training course generation method belong to the same concept. Details not described in detail in the technical solution of the training course generation device can be found in the description of the technical solution of the above-described training course generation method.

[0116] Figure 11A structural block diagram of a computing device 1100 according to an embodiment of this application is shown. The components of the computing device 1100 include, but are not limited to, a memory 1110 and a processor 1120. The processor 1120 is connected to the memory 1110 via a bus 1130, and a database 1150 is used to store data.

[0117] The computing device 1100 also includes an access device 1140, which enables the computing device 1100 to communicate via one or more networks 1160. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 1140 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0118] In one embodiment of this application, the aforementioned components of the computing device 1100 and Figure 11 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 11 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0119] The computing device 1100 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 1100 can also be a mobile or stationary server.

[0120] The processor 1120 is used to execute the following computer program / instructions, which, when executed by the processor, implement the steps of the above-described training course generation method.

[0121] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the training course generation method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the training course generation method described above.

[0122] An embodiment of this specification also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the training course generation method described above.

[0123] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the training course generation method described above. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the training course generation method described above.

[0124] An embodiment of this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the training course generation method described above.

[0125] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the training course generation method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the training course generation method described above.

[0126] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0127] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0128] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0129] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0130] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A method for generating training courses, characterized in that, include: Obtain the course-generated text; The course-generated text is input into the course generation model to obtain the training course output by the course generation model. The training course is generated based on the course structure description information, which is generated based on the action structure information and the preset action library. The action structure information is determined based on the course-generated text.

2. The method as described in claim 1, characterized in that, The training courses output by the course generation model include: The course generation text is parsed using the course generation model to obtain the action structure information; The course structure description information is generated based on the action structure information and the preset action library; A training course is generated based on the course structure description information and the action structure information.

3. The method as described in claim 2, characterized in that, The course generation model is used to parse the course-generated text to obtain the action structure information, including: The course-generated text is parsed based on preset field standard information to obtain the action type and course initial text; The course reference text is obtained by parsing the action type. An information extraction template is determined based on the action type, and action description information is generated based on the information extraction template and the course reference text. The action structure information is generated based on the action type and the action description information.

4. The method as described in claim 3, characterized in that, The action structure information is generated based on the action type and the action description information, including: At least one training stage node is determined based on the action type and the action description information; Determine at least one action loop node corresponding to each training phase node, and determine at least one action node corresponding to each action loop node. The action structure information is generated based on each action node, each action loop node, and each training phase node.

5. The method as described in claim 2, characterized in that, The action structure information includes at least one action node; The course structure description information is generated based on the action structure information and the preset action library, including: Determine the text features corresponding to each action node, and generate corresponding semantic vectors based on each text feature; Determine at least one candidate action corresponding to each semantic vector in the preset action library; Based on the candidate evaluation information, each candidate action is screened, and the target action and the target action identifier corresponding to the target action are determined. Update the corresponding action node based on the target action identifier, and generate the course structure description information.

6. The method as described in claim 2, characterized in that, A training course is generated based on the course structure description information and the action structure information, including: The course structure description information is verified based on the course compliance information to obtain the verification result; If the verification result is successful, a training course is generated based on the course structure description information and the action structure information.

7. The method as described in claim 6, characterized in that, The course compliance information includes security verification information, pacing verification information, and matching coverage information; The course structure description information is verified based on the course compliance information to obtain the verification results, including: The course structure description information is verified based on the security verification information to obtain a first verification result; If the first verification result is qualified, the course structure description information is verified according to the rhythm verification information to obtain the second verification result; If the second verification result is satisfactory, the course structure description information is verified according to the matching coverage information to obtain the verification result.

8. The method as described in claim 2, characterized in that, A training course is generated based on the course structure description information and the action structure information, including: Generate course metadata based on the action structure information; Generate a course structure based on the course structure description information; The course to be trained is generated based on the course metadata and the course structure.

9. A training course generation device, characterized in that, include: The acquisition module is configured to acquire the course-generated text. The output module is configured to input the course-generated text into the course generation model to obtain the training course output by the course generation model. The training course is generated based on the course structure description information, which is generated based on the action structure information and a preset action library. The action structure information is determined based on the course-generated text.

10. A computing device, characterized in that, include: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 8.

11. A computer-readable storage medium storing a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 8.

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

Images