Industrial safety incident type recognition method, device, equipment and storage medium

Abstract

The present application relates to a method, apparatus, device, and storage medium for identifying industrial safety event types. The method includes: obtaining a plurality of sample event data corresponding to different industrial safety event types; each sample event data having a corresponding safety event type label; for each sample event data, performing word segmentation on the industrial safety event description text in the sample event data to extract at least one keyword; screening keywords that match the event keywords from the keywords to obtain at least one target keyword, and determining key event features based on the at least one target keyword; training a safety event classification model based on the key event features corresponding to each sample event data and the corresponding industrial safety event type labels to obtain a trained safety event classification model; using the safety event classification model to perform event type identification processing on target event data. Adopting this method can improve the efficiency of identifying industrial safety event types.

Description

Technical Field

[0001] The present application relates to the field of software technology, and in particular, to a method, device, equipment, and storage medium for identifying industrial safety event types. Background Art

[0002] With the continuous development of mechanization and intelligence in the industrial production process, industrial production has been rapidly improved. However, in many production processes that require human input, due to special environments, there may be some potential safety hazards and safety incidents may occur. For some safety departments, they may receive a lot of information or documents about industrial safety incidents (hereinafter referred to as incident data).

[0003] In traditional methods, the corresponding event types are found from these incident data through manual means to complete the classification of industrial safety incidents. Classifying a large amount of incident data based on manual means will greatly increase the input of labor costs, and only based on human experience to identify the event types of industrial safety incidents, not only the labor cost is high, but also the efficiency is low. Summary of the Invention

[0004] Based on this, it is necessary to provide an industrial safety event type identification method, device, equipment, storage medium, and computer program product that can improve the efficiency of identifying industrial safety event types for the above technical problems.

[0005] In a first aspect, the present application provides an industrial safety event type identification method. The method includes:

[0006] Obtain a plurality of sample event data corresponding to different industrial safety event types; each of the sample event data has a corresponding safety event type label;

[0007] For each sample event data, perform word segmentation on the industrial safety event description text in the sample event data, and extract at least one keyword characterizing the event characteristics from the word segmentation result;

[0008] Screen keywords that match preset event keywords from the extracted keywords to obtain at least one target keyword, and determine key event characteristics according to the at least one target keyword; the event keywords are keywords with co-occurrence in the accident data of industrial safety events;

[0009] Train a safety event classification model based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed to obtain a trained safety event classification model; the trained safety event classification model is used to perform event type identification processing on the target event data of the target industrial safety event to be classified.

[0010] In one embodiment, the method further includes:

[0011] Obtaining target event data of a target industrial safety event to be classified;

[0012] Extracting target key event features from the target event data;

[0013] Using the trained safety event classification model to perform event type recognition on the target key event features to obtain the event type of the target industrial safety event.

[0014] In one embodiment, before determining multiple sample event data corresponding to different industrial safety event types, the method further includes:

[0015] Obtaining multiple initial sample event data corresponding to different industrial safety events;

[0016] Filtering out sample event data belonging to atypical industrial safety events from the multiple initial sample event data to obtain multiple sample event data belonging to typical industrial safety events.

[0017] In one embodiment, there are multiple key event features corresponding to each sample event data; training the safety event classification model to be trained based on the key event features corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data includes:

[0018] Vectorizing the multiple key event features corresponding to the same sample event data respectively, and splicing the multiple vectorized data to obtain an event feature vector corresponding to the sample event data;

[0019] In each round of iterative training, inputting the event feature vectors corresponding to each sample event data and the safety event type label corresponding to the sample event data into the safety event classification model to be trained in this round to obtain the prediction result of this round;

[0020] Calculating the loss value between the prediction result of this round and the safety event type label;

[0021] Adjusting the parameters of the safety event classification model in the direction of reducing the loss value, and taking the next round as this round to continue iterative training.

[0022] In one embodiment, the preset event keywords are preset in an event keyword set; the method further includes:

[0023] Obtaining to-be-processed event data corresponding to industrial safety events in each field;

[0024] For the to-be-processed event data, confirm the set of event feature words of the to-be-processed event data; the set of event feature words includes at least one event feature word.

[0025] Take each event feature word in the set of event feature words as a target event feature word respectively. For each target event feature word, determine the event feature words in the set of event feature words that are related to the target event feature word, and obtain the set of associated event feature words corresponding to the target event feature word.

[0026] Based on the sets of associated event feature words respectively corresponding to the respective event feature words, determine the set of event keywords.

[0027] In one embodiment, there are multiple to-be-processed event data; the multiple to-be-processed event data belong to the same field; the determining the set of event keywords based on the sets of associated event feature words respectively corresponding to the respective event feature words includes:

[0028] For the set of associated event feature words of each event feature word, calculate the respective relevance degrees of the set of associated event feature words in each of the to-be-processed event data in the same field; sum and average the respective relevance degrees to obtain the average relevance degree.

[0029] According to the average relevance degree of each set of associated event feature words, sort the multiple sets of associated event feature words.

[0030] Based on the sorting result, screen out the set of event keywords from the multiple sets of associated event feature words.

[0031] In one embodiment, the determining the event feature words in the set of event feature words that are related to the target event feature word for each target event feature word and obtaining the set of associated event feature words corresponding to the target event feature word includes:

[0032] For each target event feature word, calculate the first probability that the target event feature word and non-target event feature words appear simultaneously in the to-be-processed event data; the non-target event feature word refers to an event feature word in the set of event feature words that is different from the target event feature word.

[0033] Calculate the second probability that the target event feature word appears in the to-be-processed event data.

[0034] Calculate the third probability that the non-target event feature word appears in the to-be-processed event data.

[0035] Calculate the relevance between the target event feature word and the non-target event feature word based on the first probability, the second probability, and the third probability;

[0036] Based on the relevance, determine multiple non-target event feature words related to the target event feature word;

[0037] Based on the multiple relevant non-target event feature words and the target event feature word, generate an associated event feature word set corresponding to the target event feature word.

[0038] In a second aspect, the present application further provides an industrial safety event type recognition device. The device includes:

[0039] An extraction module, configured to obtain multiple sample event data corresponding to different industrial safety event types; each sample event data has a corresponding safety event type label; for each sample event data, perform word segmentation on the industrial safety event description text in the sample event data, and extract at least one keyword characterizing the event characteristics from the word segmentation result;

[0040] A screening module, configured to screen keywords that match preset event keywords from the extracted keywords to obtain at least one target keyword, and determine key event characteristics according to the at least one target keyword; the event keywords are keywords that have co-occurrence in the accident data of industrial safety events;

[0041] A training module, configured to train a safety event classification model based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed, and obtain a trained safety event classification model; the trained safety event classification model is used to perform event type recognition processing on the target event data of the target industrial safety event to be classified.

[0042] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor executes the steps of the above industrial safety event type recognition method.

[0043] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, and the computer program is executed by a processor to perform the steps of the above industrial safety event type recognition method.

[0044] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and the computer program is executed by a processor to perform the steps of the above industrial safety event type recognition method.

[0045] The above industrial safety event type recognition method, device, computer equipment, storage medium and computer program product obtain multiple sample event data corresponding to different industrial safety event types; each sample event data has a corresponding safety event type label. For each sample event data, the industrial safety event description text in the sample event data is segmented, and at least one keyword characterizing the event characteristics is extracted from the segmentation result. Keywords matching the preset event keywords are screened from the extracted keywords to obtain at least one target keyword, and key event characteristics are determined according to the at least one target keyword. Among them, the event keyword refers to a keyword with co-occurrence in the accident data of industrial safety events. Based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data, the safety event classification model is trained until the training is completed to obtain a trained safety event classification model. By using the trained safety event classification model, event type recognition processing is performed on the target event data of the target industrial safety event to be classified, so as to obtain the event type of the target industrial safety event. Therefore, it is no longer necessary to use manual methods to identify the event type of the target industrial safety event type, thereby improving the efficiency of industrial safety event type recognition. And since the process of event type recognition does not involve manual participation, but uses the trained safety event classification model, the influence caused by human experience differences is avoided, thus ensuring the accuracy and stability of industrial safety event type recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is an application environment diagram of the industrial safety event type recognition method in an embodiment;

[0047] Figure 2 It is a schematic flowchart of the industrial safety event type recognition method in an embodiment;

[0048] Figure 3 It is a schematic flowchart of the industrial safety event type recognition method in an embodiment;

[0049] Figure 4 It is a structural block diagram of the industrial safety event type recognition device in an embodiment;

[0050] Figure 5 It is a structural block diagram of the industrial safety event type recognition device in an embodiment;

[0051] Figure 6 It is an internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] In order to make the objectives, technical solutions, and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The industrial safety event type recognition method provided by the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal 110 communicates with the server 120 through a network. Among them, the terminal 110 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices, and the server 120 can be implemented by an independent server or a server cluster composed of multiple servers.

[0054] The terminal 110 can send multiple sample event data corresponding to different industrial safety event types to the server 120. Among them, each sample event data has a corresponding security event type label. After the server 120 obtains multiple sample event data, for each sample event data, the industrial safety event description text in the sample event data is segmented, and at least one keyword characterizing the event characteristics is extracted from the segmentation result. The server 120 filters keywords that match the preset event keywords from the extracted keywords to obtain at least one target keyword, and determines the key event characteristics according to the at least one target keyword. Among them, the event keyword refers to a keyword that has co-occurrence in the accident data of industrial safety events. The server 120 trains the security event classification model based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed, and obtains a trained security event classification model. The terminal 110 can send the target event data of the target industrial safety event to be classified to the server 120, and the server 120 can use the trained security event classification model to perform event type recognition processing on the target event data of the target industrial safety event to be classified, and send the result of the event category recognition to the terminal 110.

[0055] In one embodiment, as Figure 2 shown, an industrial safety event type recognition method is provided. In this embodiment, the method is described by taking the application of the method to the server as an example. It can be understood that the method can also be applied to the terminal, and can also be applied to a system including the terminal and the server, and is realized through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0056] S202. Obtain multiple sample event data corresponding to different industrial safety event types; each sample event data has a corresponding safety event type label; for each sample event data, perform word segmentation on the industrial safety event description text in the sample event data, and extract at least one keyword characterizing the event characteristics from the word segmentation result.

[0057] Among them, an industrial safety event is an accidental event that suddenly occurs during industrial production, harms personal safety and health, or damages equipment and facilities, or causes economic losses, or leads to the temporary or permanent termination of the original production and operation activities (including activities related to production and operation activities). Industrial safety event types can be used to characterize the event results of industrial safety events. For example, falling from a height or electric shock can both be used as industrial safety event types.

[0058] Specifically, the server obtains multiple sample event data corresponding to different industrial safety event types. Among them, each sample event data has a corresponding safety event type label. The server performs word segmentation on the industrial safety event description text in the sample event data for each sample event data, obtains multiple word segments, and further extracts at least one keyword characterizing the event characteristics from the multiple word segments.

[0059] In one embodiment, the multiple sample event data are sample data belonging to typical industrial safety events.

[0060] S204. Screen the keywords extracted that match the preset event keywords to obtain at least one target keyword, and determine the key event characteristics based on the at least one target keyword; the event keywords are keywords that have co-occurrence in the accident data of industrial safety events.

[0061] Among them, the key event characteristics are the necessary factor characteristics that constitute industrial safety events. When event keywords appear in the accident data of the same type of industrial safety events, they appear together with other event keywords.

[0062] Specifically, the server screens the keywords extracted that match the preset event keywords to obtain at least one target keyword, and determines the key event characteristics based on the at least one target keyword.

[0063] In one embodiment, the server can calculate the similarity between each extracted keyword and the preset event key, and confirm the keywords that meet the similarity requirements as target keywords based on the similarity.

[0064] In one embodiment, the server can directly determine the target keyword as the key event characteristic.

[0065] In another embodiment, the server may sort the target keywords according to the similarity degree, and select the target keywords with high similarity as the key event features.

[0066] In one embodiment, industrial safety events in the same field may have different industrial safety event types.

[0067] In another embodiment, industrial safety events in the same field have the same industrial safety event type.

[0068] In one embodiment, the server may extract a set of associated event feature words with co-occurrence for each to-be-processed event data corresponding to the industrial safety events in the target field. The target field is any one of all fields.

[0069] In another embodiment, the server may extract a set of associated event feature words with co-occurrence for all to-be-processed event data corresponding to the industrial safety events in the target field. The target field is any one of all fields.

[0070] In one embodiment, for each set of associated event feature words obtained, the server may calculate the respective relevance degrees in each to-be-processed event data in the same field. Based on the respective relevance degrees, sort the multiple sets of associated event feature words, so as to screen out a set of event keywords according to the sorting result.

[0071] In one embodiment, the server may calculate the individual probability of an event feature word appearing in the to-be-processed event data; calculate the pairwise combination probability of two event feature words appearing simultaneously in the to-be-processed event data; or may also calculate the collective combination probability of multiple event feature words appearing simultaneously in the to-be-processed event data. The server calculates the relevance degree between multiple event feature words according to at least one of the individual probability, the pairwise combination probability, and the collective combination probability.

[0072] S206. Train the safety event classification model based on the key event features corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed, so as to obtain a trained safety event classification model; the trained safety event classification model is used to perform event type recognition processing on the target event data of the to-be-classified target industrial safety event.

[0073] Specifically, the server trains the security event classification model to be trained based on the key event features corresponding to each sample event data and the industrial security event type label corresponding to the sample event data until the training is completed, and obtains the trained security event classification model. The server uses the trained security event classification model to perform event type recognition processing on the target event data of the target industrial security event to be classified. It can be understood that in this way, manual intervention is not required, but the trained security event classification model is used to improve the accuracy of industrial security event type recognition.

[0074] In one embodiment, the server may extract the target key event features from the target event data of the target industrial security event to be classified, and then use the trained security event classification model based on the target key event features.

[0075] In one embodiment, the server may adjust the parameters according to the loss value in each round of iterative training of the security event classification model to reduce the loss value.

[0076] The above industrial security event type recognition method includes obtaining multiple sample event data corresponding to different industrial security event types; each sample event data has a corresponding security event type label. For each sample event data, the industrial security event description text in the sample event data is tokenized, and at least one keyword characterizing the event characteristics is extracted from the tokenization result. Keywords that match the preset event keywords are screened from the extracted keywords to obtain at least one target keyword, and the key event features are determined according to the at least one target keyword. Among them, the event keyword refers to a keyword with co-occurrence in the accident data of industrial security events. Based on the key event features corresponding to each sample event data and the industrial security event type label corresponding to the sample event data, the security event classification model to be trained is trained until the trained security event classification model is obtained. By using the trained security event classification model, event type recognition processing is performed on the target event data of the target industrial security event to be classified to obtain the event type of the target industrial security event. Therefore, there is no need to use manual means to perform event type recognition on the target industrial security event type, thereby improving the efficiency of industrial security event type recognition. And since the process of event type recognition does not involve manual participation, but uses the trained security event classification model, the accuracy and stability of industrial security event type recognition are guaranteed.

[0077] In one embodiment, the method further includes: obtaining the target event data of the target industrial security event to be classified; extracting the target key event features from the target event data; using the trained security event classification model for event type recognition for the target key event features to obtain the event type of the target industrial security event.

[0078] Specifically, the server obtains the target event data of the target industrial security event to be classified. The server performs word segmentation on the target event data, extracts at least one keyword from the word segmentation result, filters out the keywords that match the preset event keywords from the extracted keywords to obtain at least one target keyword, and determines the target key event features according to the at least one target keyword. The server uses the trained security event classification model to perform event type recognition for the target key event features, and obtains the event type of the target industrial security event.

[0079] In this embodiment, the server extracts the target key event features for the target event data of the target industrial security event to be classified, and uses the trained security event classification model to perform event type recognition for the target key event features, which improves the accuracy of the event type recognition of the target industrial security event. And there is no need for manual participation, which improves the efficiency and stability of the event type recognition of the target industrial security event.

[0080] In one embodiment, before determining the multiple sample event data corresponding to different industrial security event types, the method further includes: obtaining multiple initial sample event data corresponding to different industrial security events; filtering out the sample event data belonging to non - typical industrial security events from the multiple initial sample event data to obtain multiple sample event data belonging to typical industrial security events.

[0081] Specifically, the server obtains multiple initial sample event data corresponding to different industrial security events. The initial sample data includes typical and non - typical sample event data. The server filters out the sample event data belonging to non - typical industrial security events from the multiple initial sample event data to obtain multiple sample event data belonging to typical industrial security events.

[0082] In this embodiment, by filtering multiple initial sample event data, multiple sample event data belonging to typical industrial security events are obtained to avoid the interference of the sample event data of non - typical industrial security events on the training process of the security event classification model, so as to improve the accuracy of the prediction of the security event classification model, and thus improve the accuracy of the event type recognition of the target industrial security event.

[0083] In one embodiment, there are multiple key event features corresponding to each sample event data; training the safety event classification model to be trained based on the key event features corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data includes: vectorizing each of the multiple key event features corresponding to the same sample event data, and concatenating the multiple vectorized data to obtain an event feature vector corresponding to the sample event data; in each round of iterative training, inputting the event feature vectors corresponding to each sample event data and the safety event type label corresponding to the sample event data into the safety event classification model to be trained in this round to obtain the prediction result of this round; calculating the loss value between the prediction result of this round and the safety event type label; adjusting the parameters of the safety event classification model in the direction of reducing the loss value, and taking the next round as this round to continue iterative training.

[0084] Specifically, there are multiple key event features corresponding to each sample event data. The server vectorizes each of the multiple key event features corresponding to the same sample event data, and concatenates the multiple vectorized data to obtain an event feature vector corresponding to the sample event data. In each round of iterative training, the server inputs the event feature vectors corresponding to each sample event data and the safety event type label corresponding to the sample event data into the safety event classification model to be trained in this round to obtain the prediction result of this round. The server calculates the loss value between the prediction result of this round and the safety event type label. The server adjusts the parameters of the safety event classification model in the direction of reducing the loss value, and takes the next round as this round to continue iterative training.

[0085] In this embodiment, there are multiple key event features corresponding to each sample event data, and the safety event classification model is trained using the multiple key event features of the same sample event data to improve the accuracy of the prediction of the safety event classification model. Moreover, during the training process, the server adjusts the parameters of the safety event classification model according to the loss value between the prediction result of this round and the safety event type label, so that the loss value gradually decreases, thereby improving the accuracy of the prediction of the safety event classification model.

[0086] In one embodiment, the preset event keywords are pre-set in the event keyword set; the method further includes: obtaining the to-be-processed event data corresponding to industrial safety events in each field; for the to-be-processed event data, confirming the event feature word set of the to-be-processed event data; the event feature word set includes at least one event feature word; taking each event feature word in the event feature word set as a target event feature word respectively, and for each target event feature word, determining the event feature words related to the target event feature word in the event feature word set, so as to obtain the associated event feature word set corresponding to the target event feature word; based on the associated event feature word sets respectively corresponding to each event feature word, determining the event keyword set.

[0087] Specifically, the server obtains the to-be-processed event data corresponding to industrial safety events in each field. For the to-be-processed event data, it confirms the event feature word set of the to-be-processed event data. Among them, the event feature word set includes at least one event feature word. The server takes each event feature word in the event feature word set as a target event feature word respectively, and for each target event feature word, determines the event feature words related to the target event feature word in the event feature word set, so as to obtain the associated event feature word set corresponding to the target event feature word. After the server obtains the associated event feature word set corresponding to each target event feature word, it determines the event keyword set based on the associated event feature word sets respectively corresponding to each event feature word.

[0088] In this embodiment, the server determines the target event feature words for the to-be-processed event data corresponding to industrial safety events in each field, and determines other event feature words related to the target event feature words, so as to obtain the associated event feature word set corresponding to the target event feature words. Further, the server determines the event keyword set from multiple associated event feature word sets. In this way, the event keyword set obtained by the server is a set of relevant event keywords and is a set of necessary factor features constituting industrial safety events. By using the necessary factor features constituting industrial safety events to train the security event classification model, the prediction accuracy can be improved.

[0089] In one embodiment, there are multiple pieces of to-be-processed event data; the multiple pieces of to-be-processed event data belong to the same field; determining the event keyword set based on the associated event feature word sets respectively corresponding to each event feature word includes: calculating the respective relevance degrees of the associated event feature word sets in each piece of to-be-processed event data in the same field for each event feature word's associated event feature word set; summing and averaging the respective relevance degrees to obtain the average relevance degree; sorting the multiple associated event feature word sets according to the average relevance degree of each associated event feature word set; and screening out the event keyword set from the multiple associated event feature word sets based on the sorting result.

[0090] Specifically, there are multiple event data to be processed. The multiple event data to be processed belong to the same field. The server calculates the relevance of each associated event feature word set in each event data to be processed in the same field for each event feature word's associated event feature word set. The server sums and averages the relevances to obtain the average relevance. The server sorts the multiple associated event feature word sets according to the average relevance of each associated event feature word set. Based on the sorting result, the server filters out the event keyword sets with high average relevance from the multiple associated event feature word sets.

[0091] In this embodiment, for each event feature word's associated event feature word set, the server calculates the relevance of the associated event feature word set in each event data to be processed in the same field; sorts the multiple associated event feature word sets based on the relevances, so as to filter out the event keyword sets with high relevance. It can be understood that the obtained event keyword sets are event keyword sets with high relevance, which further ensures that the event keywords in the set are the necessary factor features constituting industrial security events. By using the necessary factor features constituting industrial security events to train the security event classification model, the prediction accuracy can be improved.

[0092] In one embodiment, for each target event feature word, determining the event feature words in the event feature word set that are related to the target event feature word, and obtaining the associated event feature word set corresponding to the target event feature word includes: for each target event feature word, calculating the first probability that the target event feature word and non-target event feature words appear simultaneously in the event data to be processed; non-target event feature words refer to event feature words in the event feature word set that are different from the target event feature word; calculating the second probability that the target event feature word appears in the event data to be processed; calculating the third probability that the non-target event feature word appears in the event data to be processed; calculating the relevance between the target event feature word and the non-target event feature word based on the first probability, the second probability, and the third probability; determining multiple non-target event feature words related to the target event feature word based on the relevance; and generating the associated event feature word set corresponding to the target event feature word based on the multiple related non-target event feature words and the target event feature word.

[0093] Specifically, for each target event feature word, the server calculates the first probability that the target event feature word and non-target event feature words appear simultaneously in the event data to be processed. Herein, the non-target event feature words refer to the event feature words in the event feature word set that are different from the target event feature word. The server calculates the second probability that the target event feature word appears in the event data to be processed. The server calculates the third probability that the non-target event feature word appears in the event data to be processed. The server calculates the correlation degree between the target event feature word and the non-target event feature words based on the first probability, the second probability, and the third probability. The server determines multiple non-target event feature words related to the target event feature word based on the correlation degree, and generates an associated event feature word set corresponding to the target event feature word based on the multiple related non-target event feature words and the target event feature word.

[0094] In one embodiment, the server uses the PMI (Pointwise Mutual Information) method to calculate the correlation degree between the target event feature word and the non-target event feature words. The PMI algorithm is as follows:

[0095]

[0096] Where x represents the seed, and the target event feature word is used as the seed. y represents the non-target event feature word. p(x, y) represents the first probability that the target event feature word and the non-target event feature word appear simultaneously in the event data to be processed, p(x) represents the second probability that the target event feature word appears in the event data to be processed, and p(y) represents the third probability that the non-target event feature word appears in the event data to be processed. PMI(x, y) represents the pointwise mutual information value between the target event feature word and the non-target event feature word. When PMI(x, y) = 0, it indicates that there is no correlation between the target event feature word and the non-target event feature word, and when PMI(x, y) > 0, it indicates that there is a correlation between the target event feature word and the non-target event feature word.

[0097] In this embodiment, for each target event feature word, the server calculates the first probability that the target event feature word and the non-target event feature words appear simultaneously in the event data to be processed; calculates the second probability that the target event feature word appears; calculates the third probability that the non-target event feature word appears. And calculates the correlation degree between the target event feature word and the non-target event feature words based on the first probability, the second probability, and the third probability. Based on the correlation degree, an associated event feature word set corresponding to the target event feature word is determined. Therefore, it is ensured that the associated event feature word set corresponding to the target event feature word is an associated event feature word set with relevance, preparing for generating an accurate event keyword set to improve the prediction accuracy of the security event classification model.

[0098] In one embodiment, as Figure 3 shown, the server obtains a plurality of initial sample event data corresponding to different industrial security events, and filters out the sample event data belonging to atypical industrial security events from the plurality of initial sample event data, so as to obtain a plurality of sample event data belonging to typical industrial security events. For each sample event data, the server performs word segmentation on the industrial security event description text in the sample event data, and extracts a plurality of keywords from the word segmentation result; screens out the keywords that match the preset event keywords from the extracted plurality of keywords, so as to determine the key event features. The server respectively vectorizes the plurality of key event features corresponding to the same sample event data, and splices the plurality of vectorized data to obtain an event feature vector corresponding to the sample event data. The server inputs the event feature vectors of the key event features corresponding to each sample event data and the industrial security event type labels corresponding to the sample event data into the security event classification model to be trained for training until the training is completed, and a trained security event classification model is obtained. The server obtains the target event data of the target industrial security event to be classified; extracts the target key event features from the target event data; uses the trained security event classification model to perform event type recognition on the target key event features, so as to obtain the event type of the target industrial security event.

[0099] It should be understood that although the steps in the flowcharts in some embodiments of the present application are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowchart may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0100] Based on the same inventive concept, an embodiment of the present application further provides an industrial security event type recognition device for implementing the industrial security event type recognition method involved above. The implementation solution provided by this device to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the industrial security event type recognition device provided below can refer to the limitations on the industrial security event type recognition method in the above text, and will not be repeated here.

[0101] In one embodiment, as Figure 4 shown, an industrial security event type recognition device 400 is provided, including: an extraction module 402, a screening module 404, and a training module 406, wherein:

[0102] An extraction module 402 is configured to obtain a plurality of sample event data corresponding to different industrial safety event types; each sample event data has a corresponding safety event type label; for each sample event data, word segmentation is performed on the industrial safety event description text in the sample event data, and at least one keyword characterizing the event characteristics is extracted from the word segmentation result.

[0103] A screening module 404 is configured to screen keywords that match preset event keywords from the extracted keywords to obtain at least one target keyword, and determine key event characteristics according to the at least one target keyword; the event keyword is a keyword with co-occurrence in the accident data of industrial safety events.

[0104] A training module 406 is configured to train a safety event classification model based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed, and obtain a trained safety event classification model; the trained safety event classification model is configured to perform event type recognition processing on the target event data of the target industrial safety event to be classified.

[0105] In one embodiment, the industrial safety event type recognition device 400 further includes a usage module, and the usage module is configured to obtain the target event data of the target industrial safety event to be classified; extract target key event characteristics from the target event data; use the trained safety event classification model to perform event type recognition on the target key event characteristics to obtain the event type of the target industrial safety event.

[0106] In one embodiment, before determining a plurality of sample event data corresponding to different industrial safety event types, the industrial safety event type recognition device 400 further includes a preprocessing module, and the preprocessing module is configured to obtain a plurality of initial sample event data corresponding to different industrial safety events; filter out the sample event data belonging to non-typical industrial safety events from the plurality of initial sample event data to obtain a plurality of sample event data belonging to typical industrial safety events.

[0107] In one embodiment, there are multiple key event features corresponding to each sample event data; the training module 406 is further configured to vectorize the multiple key event features corresponding to the same sample event data respectively, and splice the multiple vectorized data to obtain an event feature vector corresponding to the sample event data; in each round of iterative training, the event feature vectors corresponding to the respective sample event data and the safety event type labels corresponding to the sample event data are input into the safety event classification model to be trained in this round to obtain the prediction result of this round; calculate the loss value between the prediction result of this round and the safety event type label; adjust the parameters of the safety event classification model in the direction of reducing the loss value, and take the next round as this round to continue iterative training.

[0108] As Figure 5 shown, in one embodiment, the industrial safety event type recognition device 400 further includes a feature word set acquisition module 401a and a keyword set determination module 401b, where:

[0109] The feature word set acquisition module 401a is configured to acquire the to-be-processed event data corresponding to industrial safety events in each field; for the to-be-processed event data, confirm the event feature word set of the to-be-processed event data; the event feature word set includes at least one event feature word; each event feature word in the event feature word set is respectively used as a target event feature word, and for each target event feature word, determine the event feature words related to the target event feature word in the event feature word set to obtain an associated event feature word set corresponding to the target event feature word.

[0110] The keyword set determination module 401b is configured to determine an event keyword set based on the associated event feature word sets respectively corresponding to the respective event feature words.

[0111] In one embodiment, there are multiple to-be-processed event data; the keyword set determination module 401b is further configured to calculate the respective relevance degrees of the associated event feature word sets in the respective to-be-processed event data in the same field for each associated event feature word set of each event feature word; sum and average the respective relevance degrees to obtain an average relevance degree; sort the multiple associated event feature word sets according to the average relevance degree of each associated event feature word set; based on the sorting result, screen out the event keyword set from the multiple associated event feature word sets.

[0112] In one embodiment, the feature word set acquisition module 401a is further configured to calculate, for each target event feature word, a first probability that the target event feature word and non-target event feature words co-occur in the event data to be processed; the non-target event feature words refer to event feature words in the event feature word set that are different from the target event feature words; calculate a second probability that the target event feature word appears in the event data to be processed; calculate a third probability that the non-target event feature word appears in the event data to be processed; calculate the relevance between the target event feature word and the non-target event feature words based on the first probability, the second probability, and the third probability; determine a plurality of non-target event feature words related to the target event feature word based on the relevance; and generate an associated event feature word set corresponding to the target event feature word based on the plurality of related non-target event feature words and the target event feature word.

[0113] The above industrial safety event type recognition device obtains a plurality of sample event data corresponding to different industrial safety event types; each sample event data has a corresponding safety event type label. For each sample event data, the industrial safety event description text in the sample event data is segmented, and at least one keyword characterizing the event characteristics is extracted from the segmentation result. Keywords that match the preset event keywords are screened out from the extracted keywords to obtain at least one target keyword, and key event characteristics are determined based on the at least one target keyword. The event keywords refer to keywords that have co-occurrence in the accident data of industrial safety events. Based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data, the safety event classification model is trained until the training is completed to obtain a trained safety event classification model. By using the trained safety event classification model, event type recognition processing is performed on the target event data of the target industrial safety event to be classified, so as to obtain the event type of the target industrial safety event. Therefore, there is no need to use manual means to perform event type recognition on the target industrial safety event type, thereby improving the efficiency of industrial safety event type recognition. And since the process of event type recognition does not involve manual participation but uses a trained safety event classification model, the accuracy and stability of industrial safety event type recognition are ensured.

[0114] For the specific limitations of the above industrial safety event type recognition, reference can be made to the limitations of the above industrial safety event type recognition method in the foregoing text, which will not be elaborated here. Each module in the above industrial safety event type recognition device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or can be stored in the memory in the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the above respective modules.

[0115] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structural diagram may be as shown in Figure 6 . The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store event keyword data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements an industrial security event type recognition method.

[0116] Those skilled in the art can understand that Figure 6 the structure shown in is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0117] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the steps in the above method embodiments are implemented.

[0118] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0119] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0120] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above various methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0121] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0122] The above embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. An industrial safety incident type recognition method, characterized in that, The method includes: Obtaining a plurality of sample event data corresponding to different industrial safety event types; each of the sample event data has a corresponding safety event type label; For each sample event data, performing word segmentation on the industrial safety event description text in the sample event data, and extracting at least one keyword characterizing the event characteristics from the word segmentation result; Screening keywords that match the preset event keywords from the extracted keywords to obtain at least one target keyword, and determining the key event characteristics according to the at least one target keyword; the event keywords are keywords with co-occurrence in the accident data of industrial safety events; the preset event keywords are pre-set in the event keyword set; Training a safety event classification model based on the key event characteristics corresponding to each sample event data and the industrial safety event type label corresponding to the sample event data until the training is completed, obtaining a trained safety event classification model; the trained safety event classification model is used to perform event type recognition processing on the target event data of the target industrial safety event to be classified; The method further includes: Obtaining the to-be-processed event data corresponding to industrial safety events in each field; For the to-be-processed event data, confirming the event feature word set; the event feature word set includes at least one event feature word; the to-be-processed event data is multiple; the multiple to-be-processed event data belong to the same field; Taking each event feature word in the event feature word set as a target event feature word respectively, and for each target event feature word, determining the event feature words related to the target event feature word in the event feature word set to obtain the associated event feature word set corresponding to the target event feature word; For the associated event feature word set of each event feature word, calculating the respective relevance degrees of the associated event feature word set in each to-be-processed event data under the same field; summing and averaging the respective relevance degrees to obtain the average relevance degree; Sorting the multiple associated event feature word sets according to the average relevance degree of each associated event feature word set; Based on the sorting result, screening out the event keyword set from the multiple associated event feature word sets.

2. The method according to claim 1, wherein The method further includes: Obtaining the target event data of the target industrial safety event to be classified; Extracting the target key event characteristics from the target event data; Performing event type recognition on the target key event characteristics using the trained safety event classification model to obtain the event type of the target industrial safety event.

3. The method according to claim 1, characterized in that, Before obtaining the plurality of sample event data corresponding to different industrial safety event types, the method further includes: Obtaining a plurality of initial sample event data corresponding to different industrial safety events; Filtering out the sample event data belonging to non-typical industrial safety events from the plurality of initial sample event data to obtain a plurality of sample event data belonging to typical industrial safety events.

4. The method according to claim 1, wherein There are multiple key event features corresponding to each sample event data; training the security event classification model to be trained based on the key event features corresponding to each sample event data and the industrial security event type label corresponding to the sample event data includes: Vectorize each of the multiple key event features corresponding to the same sample event data, and splice the multiple vectorized data to obtain an event feature vector corresponding to the sample event data; In each round of iterative training, input the event feature vectors corresponding to each sample event data and the security event type label corresponding to the sample event data into the security event classification model to be trained in this round to obtain the prediction result of this round; Calculate the loss value between the prediction result of this round and the security event type label; Adjust the parameters of the security event classification model in the direction of reducing the loss value, and use the next round as this round to continue iterative training.

5. The method according to claim 1, wherein Determining, for each of the target event feature words, the event feature words in the event feature word set that are related to the target event feature word to obtain the associated event feature word set corresponding to the target event feature word includes: For each of the target event feature words, calculate the first probability that the target event feature word and the non-target event feature word appear simultaneously in the event data to be processed; the non-target event feature word refers to an event feature word in the event feature word set that is different from the target event feature word; Calculate the second probability that the target event feature word appears in the event data to be processed; Calculate the third probability that the non-target event feature word appears in the event data to be processed; Calculate the correlation between the target event feature word and the non-target event feature word based on the first probability, the second probability, and the third probability; Based on the correlation, determine multiple non-target event feature words related to the target event feature word; Generate the associated event feature word set corresponding to the target event feature word based on the related multiple non-target event feature words and the target event feature word.

6. An industrial safety incident type recognition device, characterized in that, The device includes: An extraction module, configured to obtain multiple sample event data corresponding to different industrial security event types; each sample event data has a corresponding security event type label; for each sample event data, perform word segmentation on the industrial security event description text in the sample event data, and extract at least one keyword characterizing the event characteristics from the word segmentation result; A screening module, configured to screen keywords that match the preset event keywords from the extracted keywords to obtain at least one target keyword, and determine key event features according to the at least one target keyword; the event keywords are keywords that have co-occurrence in the accident data of industrial security events; the preset event keywords are pre-set in the event keyword set; A training module, configured to train a security event classification model based on the key event features corresponding to each sample event data and the industrial security event type labels corresponding to the sample event data until the training is completed, obtaining a trained security event classification model; the trained security event classification model is configured to perform event type recognition processing on the target event data of the target industrial security event to be classified; A feature word set acquisition module, configured to obtain the event data to be processed corresponding to industrial security events in each field; for the event data to be processed, confirm the set of event feature words of the event data to be processed; the set of event feature words includes at least one event feature word; the event data to be processed is multiple; the multiple pieces of event data to be processed belong to the same field; each event feature word in the set of event feature words is respectively used as a target event feature word, and for each target event feature word, determine the event feature words related to the target event feature word in the set of event feature words, obtaining the set of associated event feature words corresponding to the target event feature word; A keyword set determination module, configured to calculate the relevance of each set of associated event feature words of each event feature word in each piece of event data to be processed in the same field; sum and average the relevances to obtain an average relevance; Sort the multiple sets of associated event feature words according to the average relevance of each set of associated event feature words; Based on the sorting result, screen out the set of event keywords from the multiple sets of associated event feature words.

7. The device according to claim 6, characterized in that, The feature word set acquisition module is further configured to: For each target event feature word, calculate the first probability that the target event feature word and non-target event feature words appear simultaneously in the event data to be processed; the non-target event feature words refer to the event feature words different from the target event feature word in the set of event feature words; Calculate the second probability that the target event feature word appears in the event data to be processed; Calculate the third probability that the non-target event feature word appears in the event data to be processed; Calculate the relevance between the target event feature word and the non-target event feature word based on the first probability, the second probability, and the third probability; Based on the relevance, determine multiple non-target event feature words related to the target event feature word; Based on the related multiple non-target event feature words and the target event feature word, generate the set of associated event feature words corresponding to the target event feature word.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer program product, the computer program product includes a computer program, and when the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 5.

Images