A method and system for generating personalized training content for construction safety

Abstract

This invention discloses a method and system for generating personalized training content for construction safety, applied in the field of construction safety technology. First, training data is obtained from relevant construction safety training materials to construct a training content database. Knowledge keywords are extracted and divided into multiple knowledge point units. Personal data of construction workers to be trained is collected and analyzed from multiple dimensions to generate corresponding personalized risk tag sets. Then, the first matching degree between the set of required keywords and each knowledge point unit is calculated, and the second matching degree between the personalized risk tag set and each knowledge point unit is calculated using a mapping index. Finally, combining the first matching degree, the second matching degree, and the degree of forgetting of previously trained content, a multi-queue sorting scheme is used to filter and sort target knowledge point units, generating suitable personalized training materials. This invention generates training content based on training needs and integrates the personalized data of trainees, achieving precise education on construction safety and contributing to ensuring construction safety.

Description

Technical Field

[0001] This invention relates to the field of construction safety technology, and more specifically to a method and system for generating personalized construction safety training content. Background Technology

[0002] During construction, preventing safety accidents and ensuring the personal safety of construction workers are of paramount importance. Safety training for construction workers and strict adherence to on-site safety procedures are the two pillars of ensuring construction safety. However, current construction safety training models often employ standardized training, providing all construction workers with the same content and fixed duration of standardized safety courses. For example, Chinese patent CN118155467A discloses a power grid construction safety simulation training system. Based on virtual reality technology, it creates 3D models of equipment, environment, scenarios, and construction personnel during the construction process, breaking down the process into multiple operational tasks and steps. It guides trainees through each step in a realistic scenario, thus providing safety training for power grid construction. However, it does not consider the differences among trainees. Chinese patent CN114926152A discloses a power grid construction safety assessment and training method, system, and electronic equipment. By combining assessment and training, it allows trainees to clearly understand their knowledge. Learning operational prompts can improve training effectiveness without affecting assessment results. However, this system also uses a uniform training model. This model has significant drawbacks: it cannot effectively differentiate the unique safety risks faced by different trades (e.g., electricians, scaffolders, welders), different experience levels (e.g., novices, experienced workers), and different project stages (e.g., foundation construction, main structure construction, decoration and finishing). This leads to a disconnect between training content and actual risks, weak training relevance, insufficient trainee motivation, and ultimately, low training efficiency and unsatisfactory results, failing to achieve the fundamental goal of accident prevention. Therefore, how to provide a method and system for generating personalized training content for construction safety is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0003] In view of this, the present invention provides a method and system for generating personalized construction safety training content, which realizes precise and personalized safety education and training based on demand data and personal data.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A method for generating personalized construction safety training content includes the following steps: S1. Obtain training data from relevant construction safety training materials and establish a training content database; S2. Extract text features from the training content in the database to obtain feature words related to construction safety training as knowledge keywords. Divide the training content into multiple knowledge point units with unique identifiers based on the knowledge keywords, and construct the core keyword set and association relationship of each knowledge point unit. S3. Collect personal data of construction workers to be trained, perform multi-dimensional analysis on the personal data, calculate risk characteristic index, and generate a set of personalized risk labels for the corresponding construction workers to be trained. S4. Obtain current training demand data and extract a set of demand keywords from it. Calculate the first matching degree between the set of demand keywords and each knowledge point unit. Construct a preset mapping relationship between personalized risk tags and knowledge point units. Calculate the mapping index of the personalized risk tag set to each knowledge point unit. Use the mapping index as the second matching degree between the personalized risk tag set and the corresponding knowledge point unit. S5. Based on the first matching degree, the second matching degree, and the degree of forgetting of the training content by the construction personnel to be trained, a multi-queue sorting scheme is set to filter and sort the knowledge point units; based on the multi-queue sorting results, target knowledge point units are selected according to preset rules, and corresponding content is retrieved from the training content database in combination with the personalized risk tag set to generate personalized training materials suitable for the construction personnel to be trained.

[0005] Optionally, in S2, the text feature extraction of training content in the database to obtain knowledge keywords is specifically as follows: data preprocessing and cleaning of the training content, TF-IDF vectorization and LDA topic extraction are performed based on the processed training content, for each training material in the training content, the multiple words with the highest TF-IDF and the multiple words with the highest probability in the LDA topic are selected as candidate word set, the comprehensive weight of the candidate word set is calculated and sorted in descending order, and the candidate words with the highest ranking are selected as keywords to obtain the keyword set of the corresponding training material.

[0006] Optionally, in S2, the training content is divided into multiple knowledge point units based on knowledge keywords. Specifically, the K-Means clustering algorithm is used to cluster the keyword sets of different training materials, calculate the cosine similarity between the content, and aggregate the content with high similarity into a cluster. A descriptive label is generated for each cluster as a knowledge point unit. The same training content can be associated with multiple knowledge point units based on the different keywords it contains. The association relationship of knowledge point units is constructed through rule definition. Each knowledge point unit includes: unit ID and label, core keyword set, original training content list, and association relationship.

[0007] Optionally, S3 specifically involves: acquiring basic information, project information, construction behavior record information, and historical training and assessment data of the personnel to be trained; calculating multi-dimensional risk characteristic indices; matching the corresponding risk levels according to preset rules; generating personalized risk labels corresponding to the risk levels; and summarizing them to obtain a set of personalized risk labels. The risk characteristic indices include job risk index, experience level index, behavioral risk index, and project stage risk index.

[0008] Optionally, in S4, the calculation of the first matching degree between the set of demand keywords and each knowledge point unit is as follows: obtain the current training demand data and perform data preprocessing and cleaning; extract keywords based on the processed demand data to obtain the set of demand keywords; convert the set of demand keywords into demand vectors; calculate the semantic similarity between the demand vectors and the core keyword set vectors of each knowledge point unit; and use the semantic similarity as the first matching degree of the corresponding knowledge point unit.

[0009] Optionally, in S4, the mapping index of the personalized risk label set to each knowledge point unit is calculated as follows: a mapping relationship library between personalized risk labels and knowledge point units is pre-constructed, clarifying the associated knowledge point unit, label confidence weight, and mapping weight corresponding to each risk label; for the personalized risk label set T of the construction personnel to be trained and any knowledge point unit... Iterate through the tag set T and count the mappings to knowledge point units. Calculate the mapping index based on the number of labels N: ; In the formula, Knowledge point unit The corresponding mapping index, To map to The The confidence level of each risk label is calculated based on the risk characteristic index corresponding to the trainee. For the first Risk Labels and Knowledge Points Unit The mapping weights are preset based on the correlation between the risk labels and the accidents.

[0010] Optional, S5 specifically includes: S51. Construct three parallel queues: a priority queue, an emergency queue, and a supplementary queue. Set the admission threshold, sorting rules, and extraction ratio for each queue. S52. Select knowledge point units with a first matching degree higher than a preset first threshold as the key queue, and sort them in descending order according to the first matching degree from high to low; select knowledge point units with a second matching degree higher than a preset second threshold as the emergency queue, and sort them in descending order according to the second matching degree from high to low; select knowledge point units that the construction personnel to be trained have previously trained and whose forgetting degree is higher than a preset forgetting threshold as the supplementary queue, and sort them in descending order according to the forgetting degree from high to low. S53. Perform deduplication on the knowledge point units in the three queues, and extract the corresponding number of knowledge point units from the three queues according to the preset fixed ratio, and summarize them to form the target knowledge point unit set. S54. For each knowledge point unit in the target knowledge point unit set, retrieve the corresponding original training content from the training content database, and combine it with the personalized risk tag set of the trainees to adapt and filter the training content in a scenario-based manner, generating the final personalized training materials.

[0011] Alternatively, the formula for calculating the degree of forgetting is: ; In the formula, For trainees to review the knowledge points already covered in the training units The degree of forgetting, Knowledge point unit The time elapsed between the last training session and the current time. Knowledge point unit Historical performance records Knowledge point unit The preset difficulty level, , , The preset influence factor weights for the corresponding items.

[0012] A personalized construction safety training content generation system, applying the aforementioned personalized construction safety training content generation method, includes: The knowledge base construction module is used to obtain training data from construction safety training materials and establish a training content database; The knowledge point classification module, connected to the knowledge base construction module, is used to extract textual features from training content to obtain knowledge keywords, and to divide the training content into multiple knowledge point units based on the knowledge keywords. The personalized analysis module is used to collect construction record data of construction personnel to be trained, perform multi-dimensional analysis on the construction record data, and generate a set of personalized risk tags for the corresponding construction personnel to be trained. The demand matching module, connected to the knowledge point classification module and the personalized analysis module, is used to obtain current training demand data and extract a set of demand keywords from it, calculate the first matching degree between the set of demand keywords and each knowledge point unit, and calculate the second matching degree between the personalized risk tag set and each knowledge point unit. The content generation module connects with the demand matching module and the knowledge base construction module. It filters and sorts knowledge point units based on the first and second matching degrees, selects target knowledge point units based on the sorting results, and retrieves corresponding training materials from the training content database in combination with personalized risk tag sets.

[0013] As can be seen from the above technical solution, compared with the prior art, the present invention provides a method and system for generating personalized training content for construction safety, which has the following beneficial effects: The present invention integrates the personal data and demand data of construction personnel to generate truly targeted personalized training content; it uses data generated during the construction process to drive the generation of training content, and can dynamically adjust the training strategy according to the feedback on training effectiveness and changes in personnel behavior, so that the training content always remains effective; through targeted training, it avoids the indoctrination of irrelevant content, shortens the training time, improves the attention and acceptance of trainees, and strengthens training for high-risk behaviors and weak links, which can more effectively prevent and reduce the occurrence of safety accidents. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0015] Figure 1 This is a flowchart of the method for generating personalized construction safety training content according to the present invention. Figure 2 This is a schematic diagram of the system for generating personalized training content on construction safety according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] This invention discloses a method for generating personalized construction safety training content, such as... Figure 1 As shown, it includes the following steps: S1. Obtain training data from relevant construction safety training materials and establish a training content database; S2. Extract text features from the training content in the database to obtain feature words related to construction safety training as knowledge keywords. Divide the training content into multiple knowledge point units with unique identifiers based on the knowledge keywords, and construct the core keyword set and association relationship of each knowledge point unit. S3. Collect personal data of construction workers to be trained, perform multi-dimensional analysis on the personal data, calculate risk characteristic index, and generate a set of personalized risk labels for the corresponding construction workers to be trained. S4. Obtain current training demand data and extract a set of demand keywords from it. Calculate the first matching degree between the set of demand keywords and each knowledge point unit. Construct a preset mapping relationship between personalized risk tags and knowledge point units. Calculate the mapping index of the personalized risk tag set to each knowledge point unit. Use the mapping index as the second matching degree between the personalized risk tag set and the corresponding knowledge point unit. S5. Based on the first matching degree, the second matching degree, and the degree of forgetting of the training content by the construction personnel to be trained, a multi-queue sorting scheme is set to filter and sort the knowledge point units; based on the multi-queue sorting results, target knowledge point units are selected according to preset rules, and corresponding content is retrieved from the training content database in combination with the personalized risk tag set to generate personalized training materials suitable for the construction personnel to be trained.

[0018] In this embodiment of the invention, training materials may include industry safety standards, corporate safety procedures, typical accident case reports, and safe operation videos. Knowledge keywords refer to standardized terms conforming to industry standards, selected from the aforementioned training materials through text preprocessing, TF-IDF vectorization, and LDA topic model analysis, which can characterize the safety knowledge topics corresponding to the training content.

[0019] Furthermore, in S2, the text feature extraction of training content in the database to obtain knowledge keywords is specifically as follows: data preprocessing and cleaning of the training content, TF-IDF vectorization and LDA topic extraction are performed based on the processed training content, for each training material in the training content, the multiple words with the highest TF-IDF and the multiple words with the highest probability in the LDA topic are selected as candidate word set, the comprehensive weight of the candidate word set is calculated and sorted in descending order, and the candidate words with the highest ranking are selected as keywords to obtain the keyword set of the corresponding training material.

[0020] In this embodiment of the invention, preprocessing includes operations such as word segmentation, filtering, screening, and merging; the TF-IDF vectorization operation specifically includes: ; ; ; In the formula, For word frequency, For reverse document frequency, The number of times a word appears in a document. The total number of words in the document. Total number of documents The number of documents containing the word; if a word appears frequently in the current document but infrequently in the entire document set, its TF-IDF value is high, and it is considered a keyword of that document; each document is converted into a feature vector based on TF-IDF weights, where each dimension of the vector corresponds to a word, and the weight value represents the importance of the word to the document.

[0021] LDA topic extraction treats each document as a probability mixture of multiple topics, and each topic is composed of the probability distribution of words. LDA establishes document-topic distributions and topic-word distributions, and uses probabilistic inference methods to estimate parameters, thereby automatically identifying the most representative set of topics in a document.

[0022] The calculation of the overall weight is as follows: ; In the formula, For words w The overall weight, that is, the weight of a candidate word in the candidate word set. , , The fusion coefficient is set based on experience. For words w TF-IDF value, For words w The probability in the document topic For words w Similarity between vectors and document vectors.

[0023] Furthermore, in S2, the training content is divided into multiple knowledge point units based on knowledge keywords. Specifically, the K-Means clustering algorithm is used to cluster the keyword sets of different training materials, calculate the cosine similarity between the content, and aggregate the content with high similarity into a cluster. A descriptive label is generated for each cluster as a knowledge point unit. The same training content can be associated with multiple knowledge point units based on the different keywords it contains. The association relationship of knowledge point units is constructed through rule definition. Each knowledge point unit includes: unit ID and label, core keyword set, original training content list, and association relationship.

[0024] In this embodiment of the invention, the association relationship of the knowledge point units is specifically as follows: Strong correlation: If the keywords of two knowledge point units frequently appear in the same batch of training content, they are considered to be strongly correlated; Dependency: Based on training logic or content data definition, that is, the content of a certain knowledge point unit is trained on the basis of another knowledge point unit; Causal relationship: One knowledge point unit may lead to a related incident in another knowledge point unit.

[0025] Furthermore, S3 specifically involves: acquiring basic information, project information, construction behavior record information, and historical training and assessment data of the personnel to be trained; calculating multi-dimensional risk characteristic indices; matching corresponding risk levels according to preset rules; generating personalized risk labels corresponding to the risk levels; and summarizing them to obtain a set of personalized risk labels. The risk characteristic indices include job type risk index, experience level index, behavioral risk index, and project stage risk index.

[0026] In this embodiment of the invention, the basic information of the personnel to be trained includes: job type, date of employment, work group, certification qualifications, and training history data; project information includes: current project, project stage, work area, and tasks; construction behavior record information includes: violations, safe behaviors, personnel location, special equipment operation records, and operational compliance; the job type risk index is assigned based on the job type, and the risk index for different jobs is calculated based on historical accident data; the experience level index is assigned based on years of experience and project information participated in; the behavioral risk index is calculated based on violation frequency, high-risk area work duration, and special equipment operation compliance, wherein the violation frequency is weighted based on severity; the project stage risk index is assigned based on the current project stage.

[0027] Furthermore, in S4, the calculation of the first matching degree between the set of demand keywords and each knowledge point unit is as follows: obtain the current training demand data and perform data preprocessing and cleaning; extract keywords based on the processed demand data to obtain the set of demand keywords; convert the set of demand keywords into demand vectors; calculate the semantic similarity between the demand vectors and the core keyword set vectors of each knowledge point unit; and use the semantic similarity as the first matching degree of the corresponding knowledge point unit.

[0028] Optionally, in S4, the mapping index of the personalized risk label set to each knowledge point unit is calculated as follows: a mapping relationship library between personalized risk labels and knowledge point units is pre-constructed, clarifying the associated knowledge point unit, label confidence weight, and mapping weight corresponding to each risk label; for the personalized risk label set T of the construction personnel to be trained and any knowledge point unit... Iterate through the tag set T and count the mappings to knowledge point units. Calculate the mapping index based on the number of labels N: ; In the formula, Knowledge point unit The corresponding mapping index, To map to The The confidence level of each risk label is calculated based on the risk characteristic index corresponding to the trainee. For the first Risk Labels and Knowledge Points Unit The mapping weights are preset based on the correlation between the risk labels and the accidents. Each knowledge point unit... Get one The higher the value, the more critical and urgent the content of that unit is in correcting the person's current risks.

[0029] Furthermore, S5 specifically refers to: S51. Construct three parallel queues: a priority queue, an emergency queue, and a supplementary queue. Set the admission threshold, sorting rules, and extraction ratio for each queue. S52. Select knowledge point units with a first matching degree higher than a preset first threshold as the key queue, and sort them in descending order according to the first matching degree from high to low; select knowledge point units with a second matching degree higher than a preset second threshold as the emergency queue, and sort them in descending order according to the second matching degree from high to low; select knowledge point units that the construction personnel to be trained have previously trained and whose forgetting degree is higher than a preset forgetting threshold as the supplementary queue, and sort them in descending order according to the forgetting degree from high to low. S53. Perform deduplication on the knowledge point units in the three queues, and extract the corresponding number of knowledge point units from the three queues according to the preset fixed ratio, and summarize them to form the target knowledge point unit set. S54. For each knowledge point unit in the target knowledge point unit set, retrieve the corresponding original training content from the training content database, and combine it with the personalized risk tag set of the trainees to adapt and filter the training content in a scenario-based manner, generating the final personalized training materials.

[0030] Furthermore, the formula for calculating the degree of forgetting is: ; In the formula, For trainees to review the knowledge points already covered in the training units The degree of forgetting, Knowledge point unit The time elapsed between the last training session and the current time. Knowledge point unit Historical performance records Knowledge point unit The preset difficulty level, , , The preset influence factor weights for the corresponding items.

[0031] and Figure 1 Corresponding to the method described above, this embodiment of the invention also discloses a personalized construction safety training content generation system, which applies the above-described method for generating personalized construction safety training content, such as... Figure 2 As shown, it includes: The knowledge base construction module is used to obtain training data from construction safety training materials and establish a training content database; The knowledge point classification module, connected to the knowledge base construction module, is used to extract textual features from training content to obtain knowledge keywords, and to divide the training content into multiple knowledge point units based on the knowledge keywords. The personalized analysis module is used to collect construction record data of construction personnel to be trained, perform multi-dimensional analysis on the construction record data, and generate a set of personalized risk tags for the corresponding construction personnel to be trained. The demand matching module, connected to the knowledge point classification module and the personalized analysis module, is used to obtain current training demand data and extract a set of demand keywords from it, calculate the first matching degree between the set of demand keywords and each knowledge point unit, and calculate the second matching degree between the personalized risk tag set and each knowledge point unit. The content generation module connects with the demand matching module and the knowledge base construction module. It filters and sorts knowledge point units based on the first and second matching degrees, selects target knowledge point units based on the sorting results, and retrieves corresponding training materials from the training content database in combination with personalized risk tag sets.

[0032] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0033] Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A construction safety individualized training content generation method characterized by, Includes the following steps: S1. Obtain training data from relevant construction safety training materials and establish a training content database; S2. Extract text features from the training content in the database to obtain feature words related to construction safety training as knowledge keywords. Divide the training content into multiple knowledge point units with unique identifiers based on the knowledge keywords, and construct the core keyword set and association relationship of each knowledge point unit. S3. Collect personal data of construction workers to be trained, perform multi-dimensional analysis on the personal data, calculate risk characteristic index, and generate a set of personalized risk labels for the corresponding construction workers to be trained. S4. Obtain the current training needs data and extract the set of needs keywords from it, and calculate the first matching degree between the set of needs keywords and each knowledge point unit; Construct a pre-defined mapping relationship between personalized risk tags and knowledge point units, calculate the mapping index of the personalized risk tag set to each knowledge point unit, and use the mapping index as the second matching degree between the personalized risk tag set and the corresponding knowledge point unit. S5. Based on the first matching degree, the second matching degree, and the degree of forgetting of the training content by the construction personnel to be trained, a multi-queue sorting scheme is set to filter and sort the knowledge point units; based on the multi-queue sorting results, target knowledge point units are selected according to preset rules, and corresponding content is retrieved from the training content database in combination with the personalized risk tag set to generate personalized training materials suitable for the construction personnel to be trained.

2. The method of claim 1, wherein, In S2, the text feature extraction of training content in the database to obtain knowledge keywords is specifically as follows: data preprocessing and cleaning of the training content, TF-IDF vectorization and LDA topic extraction are performed based on the processed training content, for each training material in the training content, the multiple words with the highest TF-IDF and the multiple words with the highest probability in the LDA topic are selected as candidate word set, the comprehensive weight of the candidate word set is calculated and sorted in descending order, and the candidate words with the highest ranking are selected as keywords to obtain the keyword set of the corresponding training material.

3. The method of claim 1, wherein, In S2, the training content is divided into multiple knowledge point units based on knowledge keywords. Specifically, the K-Means clustering algorithm is used to cluster the keyword sets of different training materials, calculate the cosine similarity between the content, and aggregate the content with high similarity into a cluster. A descriptive label is generated for each cluster as a knowledge point unit. The same training content can be associated with multiple knowledge point units based on the different keywords it contains. The association relationship of knowledge point units is constructed through rule definition. Each knowledge point unit includes: unit ID and label, core keyword set, original training content list, and association relationship.

4. The method of claim 1, wherein, S3 specifically involves: acquiring basic information, project information, construction behavior records, and historical training and assessment data of the personnel to be trained; calculating multi-dimensional risk characteristic indices; matching corresponding risk levels according to preset rules; generating personalized risk labels corresponding to the risk levels; and summarizing them to obtain a set of personalized risk labels. The risk characteristic indices include job type risk index, experience level index, behavioral risk index, and project stage risk index.

5. The method for generating personalized construction safety training content according to claim 1, characterized in that, In S4, the calculation of the first matching degree between the set of demand keywords and each knowledge point unit is as follows: obtain the current training demand data and perform data preprocessing and cleaning; extract keywords based on the processed demand data to obtain the set of demand keywords; convert the set of demand keywords into demand vectors; calculate the semantic similarity between the demand vectors and the core keyword set vectors of each knowledge point unit; and use the semantic similarity as the first matching degree of the corresponding knowledge point unit.

6. The method for generating personalized construction safety training content according to claim 1, characterized in that, In S4, the mapping index of the personalized risk label set to each knowledge point unit is calculated as follows: A mapping relationship database between personalized risk labels and knowledge point units is pre-constructed, clarifying the associated knowledge point unit, label confidence weight, and mapping weight for each risk label; for the personalized risk label set T of the construction personnel to be trained and any knowledge point unit… Iterate through the tag set T and count the mappings to knowledge point units. Calculate the mapping index based on the number of labels N: ； In the formula, Knowledge point unit The corresponding mapping index, To map to The The confidence level of each risk label is calculated based on the risk characteristic index corresponding to the trainee. For the first Risk Labels and Knowledge Points Unit The mapping weights are preset based on the correlation between the risk labels and the accidents.

7. The method for generating personalized construction safety training content according to claim 1, characterized in that, S5 specifically refers to: S51. Construct three parallel queues: a priority queue, an emergency queue, and a supplementary queue. Set the admission threshold, sorting rules, and extraction ratio for each queue. S52. Select knowledge point units with a first matching degree higher than a preset first threshold as the key queue, and sort them in descending order according to the first matching degree from high to low; select knowledge point units with a second matching degree higher than a preset second threshold as the emergency queue, and sort them in descending order according to the second matching degree from high to low; select knowledge point units that the construction personnel to be trained have previously trained and whose forgetting degree is higher than a preset forgetting threshold as the supplementary queue, and sort them in descending order according to the forgetting degree from high to low. S53. Perform deduplication on the knowledge point units in the three queues, and extract the corresponding number of knowledge point units from the three queues according to the preset fixed ratio, and summarize them to form the target knowledge point unit set. S54. For each knowledge point unit in the target knowledge point unit set, retrieve the corresponding original training content from the training content database, and combine it with the personalized risk tag set of the trainees to adapt and filter the training content in a scenario-based manner, generating the final personalized training materials.

8. The method for generating personalized construction safety training content according to claim 7, characterized in that, The formula for calculating the degree of forgetting is: ； In the formula, For trainees to review the knowledge points already covered in the training units The degree of forgetting, Knowledge point unit The time elapsed between the last training session and the current time. Knowledge point unit Historical performance records Knowledge point unit The preset difficulty level, , , The preset influence factor weights for the corresponding items.

9. A personalized construction safety training content generation system, characterized in that, A method for generating personalized construction safety training content according to any one of claims 1-8, comprising: The knowledge base construction module is used to obtain training data from construction safety training materials and establish a training content database; The knowledge point classification module, connected to the knowledge base construction module, is used to extract textual features from training content to obtain knowledge keywords, and to divide the training content into multiple knowledge point units based on the knowledge keywords. The personalized analysis module is used to collect construction record data of construction personnel to be trained, perform multi-dimensional analysis on the construction record data, and generate a set of personalized risk tags for the corresponding construction personnel to be trained. The demand matching module, connected to the knowledge point classification module and the personalized analysis module, is used to obtain current training demand data and extract a set of demand keywords from it, calculate the first matching degree between the set of demand keywords and each knowledge point unit, and calculate the second matching degree between the personalized risk tag set and each knowledge point unit. The content generation module connects with the demand matching module and the knowledge base construction module. It filters and sorts knowledge point units based on the first and second matching degrees, selects target knowledge point units based on the sorting results, and retrieves corresponding training materials from the training content database in combination with personalized risk tag sets.

Images