A Text-Based Data Processing Method and System for Rail Transit Emergency Response Plans

Abstract

This invention provides a text-based method and system for processing rail transit emergency response plan data, relating to the field of data processing. The method involves: using a BERT model to perform entity recognition and relationship extraction based on the corresponding text of the emergency response plan, obtaining a structured knowledge graph of the emergency response plan and historical cases, representing the relationships between operational content, involved positions, and operational time limits; responding to the occurrence of emergency events in rail transit areas, splitting the emergency guidance data in the rail transit emergency response plan into multiple task modules according to a specified standard process; optimizing and combining multiple task modules using a genetic algorithm based on the multiple transportation areas corresponding to the emergency event and the event details of the emergency event, obtaining assigned tasks for collaborative handling of emergency events in multiple transportation areas; and decomposing each assigned task into structured operational step instructions based on the relationship patterns between operational content, involved positions, and operational time limits using the knowledge graph, according to the emergency event and designated positions.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for processing rail transit emergency response plan data based on text analysis. Background Technology

[0002] Currently, efficient handling of emergencies is crucial for ensuring passenger safety and operational order during rail transit operations. Traditional emergency plans are primarily stored in document form, a method with numerous drawbacks. Existing methods for storing and handling rail transit emergency plans are insufficient to meet the demands of increasingly complex operating environments and efficient emergency response, resulting in low overall processing efficiency for rail transit emergency data. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for processing rail transit emergency response plan data based on text analysis, so as to solve the technical problem of low overall processing efficiency of rail transit emergency data.

[0004] Firstly, this application provides a text analysis-based method for processing rail transit emergency response plan data, applied to a rail transit emergency system. The method includes:

[0005] In response to the acquisition of the corresponding text of the rail transit emergency plan, entity recognition and relationship extraction are performed based on the text using the BERT model to obtain a structured emergency plan and historical case knowledge graph of the relationship between the operational content, involved positions and operational time limits in the rail transit emergency plan.

[0006] In response to the occurrence of an emergency in the rail transit area, the emergency guidance data in the rail transit emergency plan is divided into multiple task modules according to a specified standard process. Based on the multiple transportation areas corresponding to the emergency and the event situation of the emergency, a genetic algorithm is used to optimize and combine the multiple task modules to obtain the assigned tasks for multiple transportation areas to collaboratively handle the emergency.

[0007] Based on the emergency event and the designated positions, each assigned task is broken down into structured operation step instructions according to the association pattern between the operation content, the positions involved, and the operation time limit through the emergency plan and historical case knowledge graph. Based on the structured operation step instructions, target operation step instructions that are suitable for the event type of the emergency event are automatically matched, and the target operation step instructions are sent to each of the traffic areas of the emergency event.

[0008] In one possible implementation, in response to an emergency event in a rail transit area, the emergency guidance data in the rail transit emergency plan is divided into multiple task modules according to a specified standard process. Based on the multiple transit areas corresponding to the emergency event and the event details of the emergency event, a genetic algorithm is used to optimize and combine the multiple task modules to obtain an allocation task for the collaborative handling of the emergency event by multiple transit areas, including:

[0009] In response to the occurrence of emergency events in the rail transit area, the emergency guidance data in the rail transit emergency plan is divided into multiple task modules according to a specified standard process. Each task module corresponds to a specific operation task, and there is a logical relationship between multiple task modules.

[0010] The emergency task corresponding to each of the multiple traffic areas where the emergency event occurred is treated as a chromosome, and the task module is treated as a gene.

[0011] Based on the actual situation of the emergency event, the structure of the chromosomes under the gene is continuously optimized through a genetic algorithm that initializes the population, calculates fitness, and performs crossover and mutation. This optimizes the combination and allocation of the task modules in each of the traffic areas. In each iteration of the optimization, the total time required to complete the corresponding task in each of the traffic areas under the current task allocation scheme is calculated. The value of the fitness function is determined based on the total time. Through multiple iterations, the optimal task allocation scheme corresponding to the shortest time for multiple traffic areas to collaboratively handle the emergency event is determined.

[0012] In one possible implementation, the length of the chromosome is the total number of task modules, each gene position of the chromosome corresponds to a specific task module, and the value of the gene position is the region number of the traffic area to which the corresponding task module is assigned. The structure of the chromosome under the gene is continuously optimized using a genetic algorithm that initializes the population, calculates fitness, and performs crossover and mutation based on the actual situation of the emergency event. This optimizes the combination and allocation of task modules in each traffic area. In each iteration of the optimization, the total time required to complete the corresponding task in each traffic area under the current task allocation scheme is calculated. The value of the fitness function is determined based on the total time. Through multiple iterations, the optimal task allocation scheme corresponding to the shortest time for multiple traffic areas to collaboratively handle the emergency event is determined, including:

[0013] Based on the actual situation of the emergency event, an initial population containing P chromosomes is generated, where P is the population size of the initial population.

[0014] The initial population is created by assigning a region number to each of the task modules based on each chromosome, and the initial population represents the initial set of guesses for the task allocation scheme.

[0015] For each chromosome, the chromosome is decoded using a fitness function to obtain a set of task modules that need to be executed for each traffic area;

[0016] The emergency plan and historical case knowledge graph are used to analyze the task modules assigned to each traffic area, and the execution time required for each traffic area to complete all the assigned task modules is calculated.

[0017] Select the first traffic area with the longest execution time from the execution times corresponding to the multiple traffic areas, determine the execution time corresponding to the first traffic area as the total time, and determine the negative number of the total time as the fitness function value;

[0018] The following steps are performed iteratively until the termination condition is reached, where the fitness function value no longer increases:

[0019] Based on the fitness function value corresponding to each chromosome in the initial population, select multiple first chromosomes with the highest fitness function values;

[0020] Two parent chromosomes are selected from a plurality of first chromosomes. Based on the two parent chromosomes, a crossover operation is performed with a first specified probability to generate two new daughter chromosomes. For each crossover point of the crossover operation, the portion of the daughter chromosomes after the crossover point of the parent chromosomes is exchanged, so as to explore new allocation schemes by combining the superior module allocation patterns of different parents.

[0021] The newly generated offspring chromosome is modified by mutation of some gene loci with a second specified probability to obtain an initial new chromosome, so as to avoid the algorithm getting trapped in a local optimum by reassigning some of the task modules to another traffic area; wherein the second specified probability is less than the first specified probability;

[0022] The initial new chromosome is merged with the elite chromosome with the highest fitness function value that was directly retained from the previous generation to generate a new generation of population;

[0023] During the iterative execution of the steps, after each iteration, the optimal fitness function value in the current population is determined. When the iteration termination condition of no longer increasing the optimal fitness function value is reached, the algorithm stops and outputs the optimal chromosome with the highest fitness function value in the population throughout the generations.

[0024] The optimal chromosome is then decoded to obtain the optimal task allocation scheme, which is the shortest collaborative scheme derived from historical knowledge and optimization algorithms.

[0025] In one possible implementation, in response to the acquisition of the text corresponding to the rail transit emergency plan, entity recognition and relationship extraction are performed based on the text using a BERT model to obtain a structured emergency plan and historical case knowledge graph of the operational content, involved positions, and operational time limits in the rail transit emergency plan, including:

[0026] In response to the acquisition of the text corresponding to the rail transit emergency plan, the NLP is used to perform word segmentation processing on the text according to the specified syntax and semantic rules to obtain the word segmentation processing result. Then, the BERT model is used to perform entity recognition and relation extraction on the word segmentation processing result to obtain the entity objects in the rail transit emergency plan. The entity objects include operation content, involved positions and operation time limits.

[0027] The entity relationships between the operation content, the involved positions, and the operation time limit are extracted using the BERT model to determine the required data of the involved positions and the operation time limit for each operation content.

[0028] Based on the entity objects, entity relationships, and requirement data, an emergency plan ontology model is constructed using the Protege system. The emergency plan ontology model is then used to represent the operation content, the involved positions, the operation time limits, and the entity relationships through different classes. This allows the text corresponding to the rail transit emergency plan to be organized in a structured and hierarchical manner, resulting in a structured emergency plan and historical case data.

[0029] Based on the structured emergency response plan and historical case data, a structured emergency response plan and historical case knowledge graph is generated and stored using the Neo4j graph database. Each entity object in the Neo4j graph database corresponds to a node, the entity relationships between entities are corresponding edges, and the nodes and edges form a knowledge graph.

[0030] In response to a query operation for the operation steps of the designated position in the emergency event, the corresponding target node and target edge are located in the knowledge graph of the emergency plan and historical cases through the Neo4j graph database. The target operation content, the target position involved, and the target operation time limit corresponding to the target node are obtained, as well as the target relationship between the target operation content, the target position involved, and the target operation time limit corresponding to the target edge.

[0031] In one possible implementation, the rail transit emergency system is connected to a monitoring system; the method further includes:

[0032] The monitoring system acquires passenger video data within the rail transit area.

[0033] Based on the passenger video data, passenger facial expressions and lip movements are determined through image recognition, and passenger voiceprints and speech content are determined through sound recognition.

[0034] The multiple audio segments are matched based on the passenger's facial expressions, lip movements, and voiceprints to obtain matching results, and the audio content corresponding to each passenger is determined based on the matching results.

[0035] Based on the passenger video data, semantic emotion recognition is used to determine the tone and volume of each passenger's corresponding voice content.

[0036] Based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the voice content corresponding to each passenger, dialogue matching is performed on all passengers in the rail transit area to obtain passenger dialogue groups corresponding to group dialogue content.

[0037] In response to an emergency in the rail transit area, the monitoring system determines data on passenger responses to the emergency in multiple different passenger dialogue groups.

[0038] For each group of passengers in the dialogue, the AI ​​system uses the emergency plan and historical case knowledge graph to determine the task allocation strategy based on the corresponding response data, so that the task allocation strategy is adapted to the passenger dialogue group.

[0039] In one possible implementation, after the monitoring system determines data on passenger responses to the emergency in response to an emergency in the rail transit area, the system further includes:

[0040] Training samples were determined based on data from various types of historical emergency events, multiple groups of historical passenger dialogues, and corresponding historical responses.

[0041] The initial prediction model is trained using the training samples to obtain the passenger response prediction model;

[0042] In response to a notification that a target emergency event is about to occur in the rail transit area, based on the current passenger dialogue group in the rail transit area, the passenger response prediction model is used to predict passenger response and obtain passenger response prediction results. The passenger response prediction results include the level of passenger anxiety and panic and the passenger's tendency to move direction in the time period after the occurrence of the target emergency event.

[0043] Based on the passengers' level of anxiety and panic and their preferred direction of movement, the AI ​​system uses the emergency plan and historical case knowledge graph to determine the target task allocation strategy for the current passenger dialogue group in advance, so as to pre-determine the target task allocation strategy suitable for the current passenger dialogue group.

[0044] In one possible implementation, after performing dialogue matching on all passengers in the rail transit area based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the corresponding voice content of each passenger, to obtain passenger dialogue groups corresponding to grouped dialogue content, the method further includes:

[0045] In response to the presence of a specific group of people in the current passenger dialogue group, the system determines the target location of the specific group of people in the rail transit area through image recognition based on the passenger video data and marks the specific group of people in real time.

[0046] Based on the specific type and number of people in the group and the target location, the AI ​​system determines the emergency operation priority for the current passenger dialogue group and executes the target operation step instructions according to the emergency operation priority corresponding to each of the current passenger dialogue groups.

[0047] The system sends the real-time target location and navigation route to the target location to the operator's device, enabling the operator to reach the specific group of people.

[0048] In one possible implementation, after performing dialogue matching on all passengers in the rail transit area based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the corresponding voice content of each passenger, to obtain passenger dialogue groups corresponding to grouped dialogue content, the method further includes:

[0049] Based on the passenger video data, image recognition is used to determine the inter-group density among the multiple passenger dialogue groups and the degree of personnel gathering in each passenger dialogue group.

[0050] In response to a notification that an emergency is about to occur in the rail transit area, a point of high crowd concentration is determined based on the density between groups and the degree of personnel gathering;

[0051] The monitoring system emits sound waves of a specified frequency at the point where the crowd is highly concentrated, in order to disperse the crowd and prevent stampedes in emergency situations.

[0052] Secondly, this application provides a text analysis-based rail transit emergency plan data processing system, applied to rail transit emergency systems, including:

[0053] The extraction module is used to respond to the acquisition of the text corresponding to the rail transit emergency plan, and to perform entity recognition and relationship extraction based on the text using the BERT model to obtain the structured emergency plan and historical case knowledge graph of the relationship between the operation content, the positions involved and the operation time limits in the rail transit emergency plan.

[0054] The combination module is used to respond to the occurrence of emergency events in the rail transit area. It breaks down the emergency guidance data in the rail transit emergency plan into multiple task modules according to a specified standard process. Based on the multiple traffic areas corresponding to the emergency event and the event situation of the emergency event, the genetic algorithm is used to optimize and combine the multiple task modules to obtain the assigned tasks for multiple traffic areas to collaboratively handle the emergency event.

[0055] The matching module is used to break down each assigned task into structured operation step instructions based on the relationship between operation content, involved positions and operation time limits according to the relationship between the emergency event and the designated position through the emergency plan and historical case knowledge graph. Based on the structured operation step instructions, the module automatically matches the target operation step instructions that are suitable for the event type of the emergency event, and sends the target operation step instructions to each traffic area of ​​the emergency event.

[0056] Thirdly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method described in the first aspect above.

[0057] Fourthly, this application also provides a computer-readable storage medium storing computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in the first aspect above.

[0058] This application brings the following beneficial effects:

[0059] This application provides a text analysis-based method and system for processing rail transit emergency response plan data. In response to the acquisition of the corresponding text of the rail transit emergency response plan, the system uses a BERT model to perform entity recognition and relationship extraction based on the text, obtaining a structured emergency response plan and historical case knowledge graph showing the relationships between operational content, involved positions, and operational time limits. In response to the occurrence of an emergency event in a rail transit area, the system breaks down the emergency guidance data in the rail transit emergency response plan into multiple task modules according to a specified standard process. Based on the multiple traffic areas corresponding to the emergency event and the event details of the emergency event, a genetic algorithm is used to optimize and combine the multiple task modules, resulting in assigned tasks for the collaborative processing of the emergency event by multiple traffic areas. Based on the emergency event and designated positions, each assigned task is decomposed into structured operational step instructions according to the association patterns between operational content, involved positions, and operational time limits using the emergency response plan and historical case knowledge graph. Based on these structured operational step instructions, target operational step instructions adapted to the event type of the emergency event are automatically matched, and specific operational step instructions are assigned to each traffic area of ​​the emergency event. The system sends the target operation step instructions to the relevant regions. In this solution, emergency plans are stored in a structured manner and intelligently broken down into specific steps based on a knowledge graph. When an emergency occurs, the system can quickly and automatically push appropriate handling steps, eliminating the need for staff to search for information in a large number of unstructured documents, greatly shortening the time for obtaining handling measures. Simultaneously, by using a genetic algorithm to optimize task allocation, the system achieves the shortest time for multi-regional task collaboration, avoiding resource waste and time delays caused by unreasonable task allocation. This comprehensively improves the overall efficiency of emergency response. In emergencies such as large passenger flows or equipment failures, staff can quickly act according to the steps pushed by the system, effectively reducing incident handling time and ensuring the rapid restoration of normal operation. Therefore, by using the aforementioned technologies such as knowledge graphs, natural language processing, ontology model construction, and optimization algorithms to achieve structured storage and intelligent push of digital emergency plans to assist in sudden emergency response, this solution addresses the problems existing in current rail transit emergency plans regarding storage, retrieval, reliance on experience, and multi-regional collaborative handling. It improves the efficiency and accuracy of responding to complex emergency situations, ensures the safe operation of rail transit, and enhances the overall processing efficiency of rail transit emergency data.

[0060] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0062] Figure 1 A flowchart illustrating the text analysis-based data processing method for rail transit emergency response plans provided in this application embodiment;

[0063] Figure 2 A flowchart illustrating the optimization and combination of multiple task modules using a genetic algorithm in the text analysis-based rail transit emergency plan data processing method provided in this application embodiment;

[0064] Figure 3 A schematic diagram of the structure of a rail transit emergency plan data processing system based on text analysis, provided in an embodiment of this application;

[0065] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0067] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this application, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0068] Currently, unstructured document formats make it difficult to quickly and accurately retrieve and access contingency plan content. When an emergency occurs, staff need to search through large amounts of text for relevant handling measures, which is not only time-consuming but also prone to missing key information, seriously affecting the efficiency of emergency response. For example, in emergencies such as large passenger flows or equipment failures, the duty station manager, control room operator, and station staff need to quickly obtain accurate handling guidelines, but finding the right content from lengthy documents is not easy. Secondly, relying on human experience for emergency response carries significant risks. The experience levels of different staff vary, and in the face of complex and ever-changing emergencies, misjudgments or inappropriate measures may occur. For example, for some rare equipment failures, inexperienced station staff may not be able to accurately determine the type of failure, thus delaying the response. Furthermore, existing emergency plans lack an intelligent push mechanism and cannot automatically push the best handling measures based on the characteristics of the actual emergency and the situation in each area. In actual operation, different areas may face multiple emergency events simultaneously. How to provide accurate handling suggestions to station staff in each area while ensuring the shortest possible time for multi-area task coordination is a major challenge. Furthermore, traditional contingency plans cannot effectively integrate historical data for analysis. While historical emergency events have accumulated a wealth of valuable experience and lessons, the existing contingency plan system fails to fully utilize this data. This results in an inability to effectively learn from past successes or avoid repeating mistakes when facing similar events. In conclusion, the existing methods for storing and handling rail transit emergency plans are no longer sufficient to meet the demands of an increasingly complex operating environment and efficient emergency response, leading to low overall processing efficiency for rail transit emergency data.

[0069] Based on this, the embodiments of this application provide a method and system for processing rail transit emergency plan data based on text analysis. This method can solve the technical problem of low overall processing efficiency of rail transit emergency data.

[0070] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0071] Figure 1 This is a flowchart illustrating a text analysis-based data processing method for rail transit emergency response plans, provided as an embodiment of this application. The method is applied to a rail transit emergency response system. Figure 1 As shown, the method includes:

[0072] Step S110: In response to the acquisition of the text corresponding to the rail transit emergency plan, entity recognition and relationship extraction are performed based on the text using the BERT model to obtain a structured emergency plan and historical case knowledge graph of the relationship between the operational content, involved positions and operational time limits in the rail transit emergency plan.

[0073] This step achieves structured storage of emergency response plans. Specifically, NLP techniques, such as jieba word segmentation, are used to segment the plan text. BERT is used for entity recognition and relation extraction to extract key information from the emergency response plan, including involved positions, specific operations, and operation time limits. Then, an ontology model is built using the Protege system to organize this structured information in a standardized and hierarchical manner. Finally, the structured plan and historical case knowledge graph are stored in the Neo4j graph database, making the relationships between plan information clearer and facilitating rapid retrieval and access. This structured storage method overcomes the drawbacks of traditional unstructured plans, greatly improving information retrieval efficiency.

[0074] As one possible implementation, step S110 may specifically include the following steps:

[0075] In response to the acquisition of the corresponding text of the rail transit emergency plan, the NLP is used to perform word segmentation based on the text according to the specified syntax and semantic rules to obtain the word segmentation results. The BERT model is then used to perform entity recognition and relation extraction on the word segmentation results to obtain the entity objects in the rail transit emergency plan. The entity objects include the operation content, the positions involved, and the operation time limit.

[0076] The entity relationships between operation content, involved positions, and operation time limits are extracted using the BERT model to determine the required data for the involved positions and operation time limits for each operation content. Based on the entity objects, entity relationships, and required data, an emergency plan ontology model is constructed using the Protege system. The emergency plan ontology model is then used to represent operation content, involved positions, operation time limits, and entity relationships through different classes, so as to organize the corresponding text of the rail transit emergency plan in a structured and hierarchical manner, resulting in structured emergency plans and historical case data.

[0077] Based on the structured emergency response plans and historical case data, the Neo4j graph database is used to generate and store a structured emergency response plan and historical case knowledge graph. In the Neo4j graph database, each entity object corresponds to a node, and the entity relationships between entities are corresponding edges. The nodes and edges form the knowledge graph.

[0078] In response to queries regarding the operational steps for a specific position in an emergency, the system locates the corresponding target nodes and edges in the knowledge graph of emergency plans and historical cases using the Neo4j graph database. This yields the target operation content, the relevant positions, and the target operation time limit for each target node, as well as the target relationship between the target operation content, the relevant positions, and the target operation time limit for each target edge.

[0079] For example, the data preprocessing process involves collecting emergency response plan texts and related operational documents in the rail transit field, cleaning these texts to remove noise data such as irrelevant annotations and formatting symbols, and standardizing the text format to prepare for subsequent processing.

[0080] NLP techniques (jieba word segmentation, BERT entity recognition / relation extraction) were used to perform text structuring processing on rail transit emergency plans and operational documents. For word segmentation, the jieba word segmentation tool was used to divide the preprocessed text into individual words according to Chinese grammar and semantic rules. For example, the sentence "The stationmaster on duty will activate the flow control device within 5 minutes" will be segmented into words such as "stationmaster on duty," "within," "within 5 minutes," "activate," and "flow control device." These word segmentation results form the basis for subsequent entity recognition and relation extraction.

[0081] For the entity recognition and relation extraction process, for example, the BERT model is used to perform entity recognition on the segmented text, extracting entity information such as job positions (e.g., duty station manager, control room operator, station staff), operations (e.g., activating the flow limiting device, checking equipment status), and time limits (e.g., within 5 minutes). Simultaneously, the BERT model further extracts the relationships between these entities to determine the job position corresponding to each operation and the time requirements for that operation. For example, it is clarified that the operation of "activating the flow limiting device" is to be completed by the "duty station manager" within "5 minutes".

[0082] For the process of storing contingency plans, for example, an ontology model is built using the Protecte tool, and a knowledge graph of structured contingency plans and historical cases is stored using Neo4j. The construction of this ontology model uses Protecte to build the ontology model of the emergency response plan based on entities and relationships extracted from the text. In the ontology model, different classes are defined to represent concepts such as positions, operations, and time limits, as well as the relationships between them. For example, "Position Class," "Operation Class," and "Time Class" are created, and relationships such as "Operation is performed by Position" and "Operation has a time limit" are established. In this way, the information in the emergency response plan is organized in a structured and hierarchical manner.

[0083] For storing the knowledge graph, for example, the constructed ontology model and related historical case information are stored using the Neo4j graph database. In Neo4j, each entity is a node, and the relationships between entities are edges, forming a knowledge graph. Thus, when querying the operational steps of a specific position in a particular emergency, the graph database can quickly locate the relevant nodes and edges to obtain the required information. For example, querying "operational steps of station staff under high passenger flow conditions," the system can quickly find the corresponding nodes and relationships in the knowledge graph and return the relevant information. The aforementioned ontology model can graphically present the emergency response ontology model built using Protege, displaying different categories such as "position category," "operation category," and "time category," as well as the connections between them.

[0084] In this embodiment, Protege is used to construct an ontology model and Neo4j is used to store the knowledge graph. This not only enables structured storage of emergency plans but also facilitates the organization and analysis of historical cases. Newly hired staff can quickly learn from past emergency response experiences and understand how to handle different types of events through the system. Simultaneously, as new emergency events occur and response experience accumulates, the system can continuously update and improve the knowledge graph, achieving continuous accumulation and inheritance of knowledge, further enhancing the overall level of the rail transit emergency management system, and facilitating knowledge management and experience utilization.

[0085] Step S120: In response to the occurrence of an emergency event in the rail transit area, the emergency guidance data in the rail transit emergency plan is divided into multiple task modules according to the specified standard process. Based on the multiple traffic areas corresponding to the emergency event and the event situation of the emergency event, the genetic algorithm is used to optimize and combine the multiple task modules to obtain the allocation task for the collaborative handling of the emergency event in multiple traffic areas.

[0086] In this step, a genetic algorithm is used to optimize the allocation of emergency response tasks across multiple regions, minimizing the time required for coordinated action. For example, the genetic algorithm is employed to achieve the goal of minimizing the time needed for coordinated action across multiple regions. All existing and subsequently released operational guidelines are broken down into related independent modules according to a standardized process. Based on the actual conditions of different regions and the characteristics of the emergency events, the genetic algorithm is used to optimize and combine these modules, rationally allocating tasks so that each region can achieve optimal results in coordinated emergency response, effectively improving emergency response efficiency.

[0087] In one possible implementation, such as Figure 2 As shown, step S120 may specifically include the following steps:

[0088] Step S122: In response to the occurrence of an emergency in the rail transit area, the emergency guidance data in the rail transit emergency plan is divided into multiple task modules according to the specified standard process; each task module corresponds to a specific operation task, and there are logical connections between multiple task modules.

[0089] Step S124: Treat the emergency task corresponding to each of the multiple traffic areas where the emergency event occurred as a chromosome, and treat the task module as a gene;

[0090] Step S126: Based on the actual situation of the emergency event, the structure of chromosomes under the gene is continuously optimized by initializing the population, calculating fitness, and using a genetic algorithm for crossover and mutation to optimize the combination and allocation of task modules in each traffic area. In each iteration of the optimization, the total time required to complete the corresponding task in each traffic area under the current task allocation scheme is calculated. The value of the fitness function is determined based on the total time. Through multiple iterations, the optimal task allocation scheme corresponding to the shortest time for multiple traffic areas to collaboratively handle the emergency event is determined.

[0091] For the task decomposition process, for example, existing work instructions can be broken down into independent modules according to a standardized process. Each module corresponds to a specific operational task, such as "checking the elevator's operating status" or "guiding passengers to evacuate." There are certain logical connections between these modules. For example, "guiding passengers to evacuate" may need to be performed after an anomaly is found in "checking the elevator's operating status."

[0092] For example, when using genetic algorithms, the goal is to minimize the collaborative processing time of multi-region tasks. Genetic algorithms are used to optimize the allocation of the divided task modules. First, the emergency tasks in each region are considered as chromosomes, and the task modules as genes. Then, through genetic algorithm operations such as population initialization, fitness calculation, selection, crossover, and mutation, the structure of the chromosomes—that is, the allocation method of task modules in each region—is continuously optimized. In each iteration, the total time required for each region to complete the task under the current task allocation scheme is calculated, and the total time is used as the value of the fitness function. Through multiple iterations, the optimal task allocation scheme that minimizes the collaborative processing time of multi-region tasks is found.

[0093] In one optional implementation, the length of the chromosome is the total number of task modules, each gene position of the chromosome corresponds to a specific task module, and the value of the gene position is the region number of the traffic area to which the corresponding task module is assigned. The genetic algorithm, which initializes the population, calculates fitness, and performs crossover and mutation based on the actual situation of the emergency event, continuously optimizes the structure of the chromosome under the gene to optimize the combination and allocation of task modules in each traffic area. In each iteration of the optimization, the total time required to complete the corresponding task in each traffic area under the current task allocation scheme is calculated, and the value of the fitness function is determined based on the total time. Through multiple iterations, the optimal task allocation scheme corresponding to the shortest time for multiple traffic areas to collaboratively handle the emergency event is determined. Specifically, this may include the following steps:

[0094] An initial population containing P chromosomes is generated based on the actual situation of the emergency event, where P is the population size of the initial population; each chromosome is created by assigning a region number to each task module, and the initial population represents the initial set of guesses for the task allocation scheme.

[0095] For each chromosome, the chromosome is decoded using a fitness function to obtain the set of task modules that need to be executed in each traffic area; the task modules assigned to each traffic area are analyzed using emergency plans and historical case knowledge graphs, and the execution time required for the traffic area to complete all assigned task modules is calculated;

[0096] Select the first traffic area with the longest execution time from the execution times corresponding to multiple traffic areas, determine the execution time corresponding to the first traffic area as the total time, and determine the negative of the total time as the fitness function value;

[0097] Iterative optimization is performed by repeating the following steps until the termination condition of the fitness function value no longer increases is reached: select the first chromosome with the highest fitness function value based on the fitness function value of each chromosome in the initial population;

[0098] Two parent chromosomes are selected from multiple first chromosomes. Based on the two parent chromosomes, a crossover operation is performed with a first specified probability to generate two new offspring chromosomes. For each crossover point, the offspring chromosomes after the crossover point of the parent chromosomes are exchanged. In order to explore new allocation schemes by combining the best module allocation patterns of different parents.

[0099] The newly generated offspring chromosome is modified by mutation with a second specified probability to change the values ​​of some gene loci, resulting in an initial new chromosome. This is done to avoid the algorithm getting trapped in a local optimum by reassigning some task modules to another traffic area. The second specified probability is less than the first specified probability.

[0100] The initial new chromosome is merged with the elite chromosome with the highest fitness function value directly retained from the previous generation to generate a new generation population. During the cyclic execution of the steps, after each iteration, the best fitness function value in the current population is determined. The algorithm stops when the iteration termination condition of the best fitness function value no longer increases is reached, and the optimal chromosome with the highest fitness function value in the population of all generations is output. The optimal chromosome is then finally decoded to obtain the optimal task allocation scheme, which is the shortest cooperative scheme obtained based on historical knowledge and optimization algorithms.

[0101] The fitness function described above can also be determined using the following formula:

[0102]

[0103] in: Indicates the fitness value; This represents the total time taken for all regions to complete their assigned tasks. This indicates a penalty for violating time constraints. For example, certain tasks must be completed within a time limit, and if the time limit is exceeded, a penalty will be added. This indicates a penalty for violating resource constraints. For example, if a position is assigned multiple conflicting tasks at the same time, an additional penalty will be imposed. and This indicates that the penalty coefficient is used to balance time and resource constraints.

[0104] Furthermore, the selection process in the above genetic algorithm can be implemented using the following formula via roulette wheel selection:

[0105]

[0106] in, Indicates the first i The probability of an individual being selected; Indicates the first i The fitness value of an individual's chromosome; This represents the fitness value of the chromosome of the j-th individual; N Indicates population size.

[0107] In this embodiment, in a scenario of collaborative emergency response across multiple traffic zones, a genetic algorithm is used to automatically optimize the allocation scheme of task modules among various zones, minimizing the maximum time required to complete tasks in all zones (i.e., the overall system completion time), thereby maximizing emergency response efficiency. By encoding the task allocation problem as a chromosome, with each gene representing the traffic zone number to which a task module is assigned, and using the longest task completion time (i.e., critical path time) across all traffic zones as the total time, and setting its inverse as the fitness function value, this directly reflects the optimization objective of minimizing the maximum completion time. This achieves globally optimal collaborative scheduling of multi-zone task allocation for emergency events, significantly improving the overall timeliness and resource utilization efficiency of emergency response.

[0108] Step S130: Based on the emergency event and designated positions, each assigned task is broken down into structured operation step instructions according to the relationship pattern between operation content, involved positions and operation time limit through emergency plan and historical case knowledge graph. Based on the structured operation step instructions, target operation step instructions that are suitable for the event type of the emergency event are automatically matched, and target operation step instructions are sent to each traffic area of ​​the emergency event.

[0109] As one possible implementation, in the intelligent decomposition of emergency plans, knowledge graph technology is used to break down the emergency plan into structured steps according to the association pattern of "job operation time limits." For example, there is a clear instruction step such as "Station manager on duty: activate the flow control device within 5 minutes." When an emergency occurs, the system can automatically match and push appropriate steps based on the type of event, such as large passenger flow or equipment failure, avoiding over-reliance on human experience and ensuring that it can provide rapid and accurate handling guidance in emergency situations.

[0110] For the specific process of sending target operation instructions to various traffic areas in response to emergency events, an example is the integration with a Closed-Circuit Television (CCTV) system to achieve automatic triggering. A selectable guidance interface based on job permissions is developed using Vue: On one hand, the CCTV system automatically triggers relevant emergency plans and pushes the best handling measures when a specific emergency scenario is detected. On the other hand, a push interface developed using Vue pushes corresponding handling suggestions and guidance to personnel in different positions, such as duty station managers, control room operators, and station staff, based on their job permissions. This allows for selection and convenient operation by staff. This automatic / manual triggering mechanism for pushing optimal handling measures by area makes emergency response more intelligent and precise, achieving intelligent triggering and push.

[0111] The specific triggering and push process can include the following methods: Automatic Triggering: The system interfaces with the CCTV system, using image recognition technology to monitor the situation within the rail transit area in real time. When a scenario matching a specific emergency event is detected, such as a sudden surge in passenger flow or an abnormal equipment malfunction, the system automatically identifies the event type and triggers the corresponding emergency plan. For example, when the CCTV system detects that the number of passengers on the platform exceeds a set threshold, it automatically determines it as a large passenger flow event and triggers the large passenger flow emergency plan. Manual Triggering: Staff can also manually request the triggering of emergency plans through the system interface based on the actual situation. For example, when a station attendant discovers an equipment malfunction that has not been promptly identified by the CCTV system, they can manually select the corresponding equipment malfunction emergency plan through the system interface. Push Interface Development: A user push interface is developed using the Vue framework. Based on the permissions of different positions, corresponding emergency response measures are pushed to the station manager, control room operator, station attendant, etc. The interface presents relevant suggestions and guidance in a selection manner for easy operation by staff. For example, for station staff, the interface displays options such as "Guide passengers to designated evacuation routes (Yes / No)" and "Assist passengers in using the emergency help button (Yes / No)." Staff simply select the appropriate option based on the situation to follow the system's guidance for emergency response. By enhancing the intelligence and personalization of the contingency plan, the system enables automatic / manual triggering of optimal response measures for different areas. Based on the actual situation and event characteristics of each area, the system accurately pushes the most suitable response suggestions to station staff in each area, and supports presentation of guidance in a selectable manner, conforming to staff operating habits and providing convenience and speed. This intelligent and personalized push mechanism allows the emergency plan to better adapt to complex and ever-changing emergency scenarios, improving the ability to cope with various complex situations.

[0112] In this embodiment, by storing emergency plans in a structured manner and intelligently decomposing them into specific steps based on a knowledge graph, the system can quickly and automatically push appropriate handling steps when an emergency occurs. This eliminates the need for staff to search for information in a large number of unstructured documents, greatly shortening the time required to obtain handling measures. Simultaneously, by using a genetic algorithm to optimize task allocation, the system achieves the shortest possible time for multi-regional task collaboration, avoiding resource waste and time delays caused by unreasonable task allocation. This comprehensively improves the overall efficiency of emergency response. In emergencies such as large passenger flows or equipment failures, staff can quickly act according to the steps pushed by the system, effectively reducing incident handling time and ensuring the rapid restoration of normal operational order. Therefore, by using the aforementioned technologies such as knowledge graphs, natural language processing, ontology model construction, and optimization algorithms to achieve structured storage and intelligent push of digital emergency plans to assist in sudden emergency response, this system solves the problems existing in current rail transit emergency plans regarding storage, retrieval, reliance on experience, and multi-regional collaborative handling. It improves the efficiency and accuracy of responding to complex emergency situations, ensures the safe operation of rail transit, and enhances the overall processing efficiency of rail transit emergency data.

[0113] Furthermore, it avoids the risks associated with over-reliance on human experience. Because the system automatically pushes out response measures based on structured plans and accurate event identification, it reduces errors caused by insufficient staff experience or misjudgment. For example, for complex equipment malfunctions, the system can accurately push out standard handling procedures for that type of malfunction, ensuring that staff take the correct measures, improving the accuracy of emergency response, reducing the possibility of the incident worsening due to incorrect handling, and enhancing the accuracy of emergency response.

[0114] In some embodiments, the rail transit emergency system is connected to a monitoring system; the method may also include the following steps:

[0115] The system acquires passenger video data in the rail transit area through a monitoring system; based on the passenger video data, it determines passenger facial expressions and lip movements through image recognition, and determines passenger voiceprints and speech content through voice recognition.

[0116] The system matches multiple audio segments based on passenger facial expressions, lip movements, and voiceprints to obtain matching results, and then determines the audio content corresponding to each passenger based on the matching results.

[0117] Based on passenger video data, semantic emotion recognition is used to determine the tone and volume of each passenger's speech content. Based on the distance between passengers, tone and volume of speech in the passenger video data, and the speech content of each passenger, dialogue matching is performed on all passengers in the rail transit area to obtain passenger dialogue groups corresponding to grouped dialogue content.

[0118] In response to emergency events in the rail transit area, the monitoring system determines the response data of passengers in multiple different passenger dialogue groups to the emergency events. For each passenger dialogue group, based on their respective response data, the AI ​​system uses emergency plans and historical case knowledge graphs to determine the corresponding task allocation strategy, so that the task allocation strategy is adapted to the passenger dialogue group.

[0119] When an emergency occurs in rail transit, multimodal perception (video combined with audio) and AI semantic understanding are used to identify and cluster the dialogue behavior and emotional reactions of passenger groups in real time. Based on the group characteristics, an adaptive emergency task allocation strategy is generated, which improves the accuracy and humanization of emergency response. This achieves a leap from passive monitoring to proactive understanding of group intentions. Emergency task allocation is no longer based solely on static area or equipment status, but dynamically adapts to the behavior and emotional characteristics of real passenger groups, thereby significantly improving the pertinence, efficiency and humanistic care of emergency response.

[0120] In some embodiments, after determining the passenger response data in response to an emergency event in a rail transit area via a monitoring system, the method may further include the following steps:

[0121] Training samples are determined based on various types of historical emergency events, multiple groups of historical passenger dialogues, and corresponding historical response data; the initial prediction model is trained using the training samples to obtain the passenger response prediction model.

[0122] In response to a notification that a target emergency event is about to occur in the rail transit area, based on the current passenger dialogue group in the rail transit area, a passenger response prediction model is used to predict passenger response, and the passenger response prediction results are obtained. The passenger response prediction results include the level of passenger anxiety and panic and the passenger's tendency to move direction in the time period after the occurrence of the target emergency event.

[0123] Based on the level of passenger anxiety and panic and the direction of passenger movement, the AI ​​system uses emergency plans and historical case knowledge graphs to determine the target task allocation strategy for the current passenger dialogue group in advance, so as to determine the target task allocation strategy that is suitable for the current passenger dialogue group.

[0124] By integrating group dialogue understanding with machine learning prediction, this system enables predictive, personalized, and proactive task allocation for future emergencies, significantly improving the resilience and service level of the rail transit system during unforeseen events. Before an actual rail transit emergency occurs, based on the current passenger group's dialogue behavior characteristics, a trained predictive model is used to anticipate their emotional state and movement tendencies, and dynamically generates adaptive emergency task allocation strategies accordingly, thereby achieving intelligent emergency dispatching with pre-emptive intervention and proactive response.

[0125] In some embodiments, after performing dialogue matching on all passengers in the rail transit area based on the distance between passengers, tone of voice, volume of voice, and the corresponding voice content of each passenger in the passenger video data to obtain passenger dialogue groups corresponding to the grouped dialogue content, the method may further include the following steps:

[0126] In response to the presence of a specific group within the current passenger chat group, the system uses image recognition based on passenger video data to determine the target location of the specific group in the rail transit area and marks the specific group in real time.

[0127] Based on the specific type, number, and target location of a particular group of people, the AI ​​system determines the emergency operation priority for the current passenger dialogue group and executes the target operation steps according to the emergency operation priority corresponding to each of the multiple current passenger dialogue groups; it also sends the real-time target location and navigation route to the target location to the operator so that the operator can reach the specific group of people.

[0128] In rail transit emergency scenarios, the system automatically identifies and locates specific groups of people (such as the elderly, children, people with disabilities, pregnant women, etc.) in real time. Combined with the context information of their respective dialogue groups, it dynamically generates differentiated emergency operation priorities and guides on-site operators to accurately and efficiently reach the target location to carry out rescue. This significantly improves the pertinence and timeliness of emergency response, enabling the system to see, accurately assess, locate, and quickly rescue specific vulnerable groups in rail transit, and builds a proactive and precise emergency rescue capability for high-risk individuals.

[0129] In some embodiments, after performing dialogue matching on all passengers in the rail transit area based on the distance between passengers, tone of voice, volume of voice, and the corresponding voice content of each passenger in the passenger video data to obtain passenger dialogue groups corresponding to the grouped dialogue content, the method may further include the following steps:

[0130] Based on passenger video data, image recognition is used to determine the density between multiple passenger dialogue groups and the degree of crowd gathering within each passenger dialogue group; in response to notification of an impending emergency in the rail transit area, the density between groups and the degree of crowd gathering are used to determine the points of high crowd concentration; the monitoring system emits sound waves of a specified frequency at these high crowd concentration points to disperse the crowd and prevent stampedes during emergencies.

[0131] By combining AI visual perception with directional acoustic intervention, an integrated stampede risk prevention and control system is constructed, encompassing identification, early warning, and proactive crowd evacuation. This system provides proactive and automated protection for crowd safety in rail transit scenarios. Before a rail transit emergency occurs, visual intelligence analyzes the spatial distribution characteristics of the crowd to accurately identify high-risk areas for stampedes. Proactive directional acoustic technology is then used for non-contact and non-intrusive crowd evacuation, effectively preventing stampede accidents.

[0132] Figure 3 A schematic diagram of a text-based rail transit emergency response plan data processing system is provided. This device can be applied to rail transit emergency response systems. Figure 3 As shown, the rail transit emergency response plan data processing system 300 based on text analysis includes:

[0133] The extraction module 301 is used to respond to the acquisition of the text corresponding to the rail transit emergency plan, and to perform entity recognition and relationship extraction based on the text using the BERT model to obtain the structured emergency plan and historical case knowledge graph of the relationship between the operation content, the positions involved and the operation time limits in the rail transit emergency plan.

[0134] The combination module 302 is used to respond to the occurrence of emergency events in the rail transit area, to divide the emergency guidance data in the rail transit emergency plan into multiple task modules according to a specified standard process, and to optimize and combine the multiple task modules using a genetic algorithm based on the multiple traffic areas corresponding to the emergency event and the event situation of the emergency event, so as to obtain the assigned task of multiple traffic areas coordinating the handling of the emergency event.

[0135] The matching module 303 is used to break down each assigned task into structured operation step instructions according to the association pattern between operation content, involved positions and operation time limits based on the emergency event and the designated position through the emergency plan and historical case knowledge graph. Based on the structured operation step instructions, it automatically matches the target operation step instructions that are suitable for the event type of the emergency event, and sends the target operation step instructions to each traffic area of ​​the emergency event.

[0136] The rail transit emergency plan data processing system based on text analysis provided in this application has the same technical features as the rail transit emergency plan data processing method based on text analysis provided in the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0137] An electronic device provided in this application embodiment, such as Figure 4 As shown, the electronic device 400 includes a processor 402 and a memory 401. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the method provided in the above embodiments.

[0138] See Figure 4 The electronic device also includes a bus 403 and a communication interface 404. The processor 402, the communication interface 404 and the memory 401 are connected via the bus 403. The processor 402 is used to execute executable modules, such as computer programs, stored in the memory 401.

[0139] The memory 401 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 404 (which can be wired or wireless), such as the Internet, wide area network, local area network, or metropolitan area network.

[0140] Bus 403 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0141] The memory 401 is used to store programs. After receiving an execution instruction, the processor 402 executes the program. The method executed by the apparatus defined by the process disclosed in any of the preceding embodiments of this application can be applied to the processor 402 or implemented by the processor 402.

[0142] Processor 402 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 402 or by instructions in software form. The processor 402 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 401, and processor 402 reads the information from memory 401 and, in conjunction with its hardware, completes the steps of the above method.

[0143] Corresponding to the above-described text-based rail transit emergency plan data processing method, this application embodiment also provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are invoked and executed by a processor, the computer-executable instructions cause the processor to perform the steps of the above-described text-based rail transit emergency plan data processing method.

[0144] The text analysis-based rail transit emergency response plan data processing system provided in this application embodiment can be specific hardware on the device or software or firmware installed on the device. The device provided in this application embodiment has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0145] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0146] For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0147] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0148] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0149] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the text analysis-based rail transit emergency plan data processing method described in the various embodiments of this application. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0150] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0151] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for processing rail transit emergency response plan data based on text analysis, characterized in that, The method, applied to rail transit emergency systems, includes: In response to the acquisition of the corresponding text of the rail transit emergency plan, entity recognition and relationship extraction are performed based on the text using the BERT model to obtain a structured emergency plan and historical case knowledge graph of the relationship between the operational content, involved positions and operational time limits in the rail transit emergency plan. In response to an emergency event in a rail transit area, the emergency guidance data in the rail transit emergency plan is broken down into multiple task modules according to a specified standard process. Each task module corresponds to a specific operational task, and there are logical connections between multiple task modules. The emergency tasks corresponding to each of the multiple rail transit areas where the emergency event occurs are treated as a chromosome, and the task modules are treated as genes. The length of the chromosome is the total number of task modules, and each gene position of the chromosome corresponds to a specific task module. The value of the gene position is the area number of the rail transit area to which the corresponding task module is assigned. Based on the actual situation of the emergency event, a P-value sequence is generated. The initial population of the chromosome, where P is the population size of the initial population; created based on each chromosome by assigning a region number to each task module, the initial population representing the initial guess set of the task allocation scheme; decoding each chromosome using a fitness function to obtain the set of task modules that need to be executed in each traffic area; analyzing the task modules assigned to each traffic area using the emergency plan and historical case knowledge graph, and calculating the execution time required for the traffic area to complete all assigned task modules; selecting the first traffic area with the longest execution time from the execution times corresponding to multiple traffic areas, and assigning the first... The execution time corresponding to the traffic area is determined as the total time, and the negative of the total time is determined as the fitness function value. Iterative optimization is performed through the following steps until a termination condition is reached where the fitness function value no longer increases: Based on the fitness function value corresponding to each chromosome in the initial population, select multiple first chromosomes with the highest fitness function values; select two parent chromosomes from the multiple first chromosomes; perform a crossover operation on the two parent chromosomes with a first specified probability to generate two new child chromosomes; and for each crossover point, exchange the portion of the child chromosomes after the parent chromosome crossover point to combine the best module allocation patterns of different parent chromosomes. Explore new allocation schemes; modify the values ​​of some gene loci through mutation on the newly generated offspring chromosomes with a second specified probability to obtain an initial new chromosome, so as to avoid the algorithm getting trapped in a local optimum by reassigning some of the task modules to another traffic area; wherein, the second specified probability is less than the first specified probability; merge the initial new chromosome with the elite chromosome with the highest fitness function value directly retained in the previous generation to generate a new generation population; after each iteration during the cyclic execution of the steps, determine the best fitness function value in the current population; when the iteration termination condition of the best fitness function value no longer increases is reached, the algorithm stops and outputs the best chromosome with the highest fitness function value in the population of all generations;The optimal chromosome is finally decoded to obtain the optimal task allocation scheme, which is the shortest collaborative scheme based on historical knowledge and optimization algorithm. Based on the emergency event and the designated positions, each assigned task is broken down into structured operation step instructions according to the relationship pattern between the operation content, the positions involved, and the operation time limit through the emergency plan and historical case knowledge graph. Based on the structured operation step instructions, target operation step instructions that are suitable for the event type of the emergency event are automatically matched, and the target operation step instructions are sent to each of the traffic areas of the emergency event.

2. The method according to claim 1, characterized in that, The process involves obtaining the corresponding text of the rail transit emergency plan, and then using a BERT model to perform entity recognition and relationship extraction based on the text. This yields a structured emergency plan and historical case knowledge graph, including the relationships between operational content, involved positions, and operational time limits within the rail transit emergency plan. In response to the acquisition of the text corresponding to the rail transit emergency plan, the NLP is used to perform word segmentation processing on the text according to the specified syntax and semantic rules to obtain the word segmentation processing result. Then, the BERT model is used to perform entity recognition and relation extraction on the word segmentation processing result to obtain the entity objects in the rail transit emergency plan. The entity objects include operation content, involved positions and operation time limits. The entity relationships between the operation content, the involved positions, and the operation time limit are extracted using the BERT model to determine the required data of the involved positions and the operation time limit for each operation content. Based on the entity objects, entity relationships, and requirement data, an emergency plan ontology model is constructed using the Protege system. The emergency plan ontology model is then used to represent the operation content, the involved positions, the operation time limits, and the entity relationships through different classes. This allows the text corresponding to the rail transit emergency plan to be organized in a structured and hierarchical manner, resulting in a structured emergency plan and historical case data. Based on the structured emergency response plan and historical case data, a structured emergency response plan and historical case knowledge graph is generated and stored using the Neo4j graph database. Each entity object in the Neo4j graph database corresponds to a node, the entity relationships between entities are corresponding edges, and the nodes and edges form a knowledge graph. In response to a query operation for the operation steps of the designated position in the emergency event, the corresponding target node and target edge are located in the knowledge graph of the emergency plan and historical cases through the Neo4j graph database. The target operation content, the target position involved, and the target operation time limit corresponding to the target node are obtained, as well as the target relationship between the target operation content, the target position involved, and the target operation time limit corresponding to the target edge.

3. The method according to claim 1, characterized in that, The rail transit emergency system is connected to a monitoring system; the method also includes: The monitoring system acquires passenger video data within the rail transit area. Based on the passenger video data, passenger facial expressions and lip movements are determined through image recognition, and passenger voiceprints and speech content are determined through sound recognition. The multiple audio segments are matched based on the passenger's facial expressions, lip movements, and voiceprints to obtain matching results, and the audio content corresponding to each passenger is determined based on the matching results. Based on the passenger video data, semantic emotion recognition is used to determine the tone and volume of each passenger's corresponding voice content. Based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the voice content corresponding to each passenger, dialogue matching is performed on all passengers in the rail transit area to obtain passenger dialogue groups corresponding to group dialogue content. In response to an emergency in the rail transit area, the monitoring system determines data on passenger responses to the emergency in multiple different passenger dialogue groups. For each group of passengers in the dialogue, the AI ​​system uses the emergency plan and historical case knowledge graph to determine the task allocation strategy based on the corresponding response data, so that the task allocation strategy is adapted to the passenger dialogue group.

4. The method according to claim 3, characterized in that, In response to an emergency event occurring in the rail transit area, after the monitoring system determines data on passenger responses to the emergency event across multiple different groups of passenger dialogue groups, the system further includes: Training samples were determined based on data from various types of historical emergency events, multiple groups of historical passenger dialogues, and corresponding historical responses. The initial prediction model is trained using the training samples to obtain the passenger response prediction model; In response to a notification that a target emergency event is about to occur in the rail transit area, based on the current passenger dialogue group in the rail transit area, the passenger response prediction model is used to predict passenger response and obtain passenger response prediction results. The passenger response prediction results include the level of passenger anxiety and panic and the passenger's tendency to move direction in the time period after the occurrence of the target emergency event. Based on the passengers' level of anxiety and panic and their preferred direction of movement, the AI ​​system uses the emergency plan and historical case knowledge graph to determine the target task allocation strategy for the current passenger dialogue group in advance, so as to pre-determine the target task allocation strategy suitable for the current passenger dialogue group.

5. The method according to claim 4, characterized in that, After performing dialogue matching on all passengers in the rail transit area based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the corresponding voice content of each passenger to obtain passenger dialogue groups corresponding to grouped dialogue content, the method further includes: In response to the presence of a specific group of people in the current passenger dialogue group, the system determines the target location of the specific group of people in the rail transit area through image recognition based on the passenger video data and marks the specific group of people in real time. Based on the specific type and number of people in the group and the target location, the AI ​​system determines the emergency operation priority for the current passenger dialogue group and executes the target operation step instructions according to the emergency operation priority corresponding to each of the current passenger dialogue groups. The system sends the real-time target location and navigation route to the target location to the operator's device, enabling the operator to reach the specific group of people.

6. The method according to claim 4, characterized in that, After performing dialogue matching on all passengers in the rail transit area based on the distance between passengers in the passenger video data, the tone of voice, the volume of voice, and the corresponding voice content of each passenger to obtain passenger dialogue groups corresponding to grouped dialogue content, the method further includes: Based on the passenger video data, image recognition is used to determine the inter-group density among the multiple passenger dialogue groups and the degree of personnel gathering in each passenger dialogue group. In response to a notification that an emergency is about to occur in the rail transit area, a point of high crowd concentration is determined based on the density between groups and the degree of personnel gathering; The monitoring system emits sound waves of a specified frequency at the point where the crowd is highly concentrated, in order to disperse the crowd and prevent stampedes in emergency situations.

7. A data processing system for rail transit emergency response plans based on text analysis, characterized in that, Applications in rail transit emergency systems include: The extraction module is used to respond to the acquisition of the text corresponding to the rail transit emergency plan, and to perform entity recognition and relationship extraction based on the text using the BERT model to obtain the structured emergency plan and historical case knowledge graph of the relationship between the operation content, the positions involved and the operation time limits in the rail transit emergency plan. The combination module is used to respond to the occurrence of emergency events in the rail transit area. It breaks down the emergency guidance data in the rail transit emergency plan into multiple task modules according to a specified standard process. Based on the multiple traffic areas corresponding to the emergency event and the event situation of the emergency event, the genetic algorithm is used to optimize and combine the multiple task modules to obtain the assigned tasks for multiple traffic areas to collaboratively handle the emergency event. The matching module is used to break down each assigned task into structured operation step instructions according to the relationship pattern between operation content, involved positions and operation time limit based on the emergency event and the designated position through the emergency plan and historical case knowledge graph. Based on the structured operation step instructions, the module automatically matches the target operation step instructions that are suitable for the event type of the emergency event and sends the target operation step instructions to each traffic area of ​​the emergency event. The combined module is specifically used for: responding to the occurrence of an emergency event in the rail transit area, splitting the emergency guidance data in the rail transit emergency plan into multiple task modules according to a specified standard process, wherein each task module corresponds to a specific operation task, and there is a logical relationship between the multiple task modules; treating the emergency task corresponding to each of the multiple transportation areas where the emergency event occurs as a chromosome, and treating the task module as a gene; continuously optimizing the structure of the chromosome under the gene through a genetic algorithm that initializes the population, calculates fitness, and performs crossover and mutation based on the actual situation of the emergency event, so as to optimize the combination and allocation method of the task modules in each of the transportation areas, and in each iteration of the optimization, calculating the total time required to complete the corresponding task in each of the transportation areas under the current task allocation scheme, determining the value of the fitness function based on the total time, and determining the optimal task allocation scheme corresponding to the shortest time for multiple transportation areas to collaboratively handle the emergency event through multiple iterations; The length of the chromosome is the total number of the task modules. Each gene position of the chromosome corresponds to a specific task module, and the value of the gene position is the region number of the traffic area to which the corresponding task module is assigned. The combination module is further configured to: generate an initial population containing P chromosomes based on the actual situation of the emergency event, where P is the population size of the initial population; create a population based on each chromosome by assigning a region number to each task module, where the initial population represents the initial guess set of the task allocation scheme; and decode each chromosome using a fitness function to obtain each traffic area. The set of task modules that the domain needs to execute; analyze the task modules assigned to each traffic area through the emergency plan and historical case knowledge graph, and calculate the execution time required for the traffic area to complete all the assigned task modules; select the first traffic area with the longest execution time from the execution times corresponding to multiple traffic areas, determine the execution time corresponding to the first traffic area as the total time, and determine the negative of the total time as the fitness function value; perform iterative optimization through the cyclic execution of the following steps until the termination condition of the fitness function value no longer increases is reached: based on each chromosome corresponding to the initial population The fitness function value is selected from a plurality of first chromosomes with the highest fitness function value; two parent chromosomes are selected from the plurality of first chromosomes, and a crossover operation is performed on the two parent chromosomes with a first specified probability to generate two new offspring chromosomes. For each crossover point of the parent chromosomes, a portion of the offspring chromosomes after the crossover point is swapped to explore new allocation schemes by combining the superior module allocation patterns of different parents; the newly generated offspring chromosomes are modified by mutation with a second specified probability to change the values ​​of some gene loci to obtain an initial new chromosome, in order to avoid computation by reassigning some of the task modules to another traffic area. The algorithm gets stuck in a local optimum; wherein the second specified probability is less than the first specified probability; the initial new chromosome is merged with the elite chromosome with the highest fitness function value directly retained from the previous generation to generate a new generation population; after each iteration during the cyclic execution of the steps, the best fitness function value in the current population is determined, and the algorithm stops when the iteration termination condition of the best fitness function value no longer increases is reached, and the optimal chromosome with the highest fitness function value in the population of all generations is output; the optimal chromosome is finally decoded to obtain the optimal task allocation scheme, which is the shortest time collaborative scheme obtained based on historical knowledge and optimization algorithm.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 6.

Images