Method and system for rapidly identifying wounded information based on artificial intelligence

Abstract

The invention provides a wounded person information rapid identification method and system based on artificial intelligence. The method comprises the following steps: receiving a multi-modal data stream through a plurality of on-site rescue devices, carrying out event association processing, and generating a comprehensive data fragment set; based on the comprehensive data fragment set, performing deep semantic analysis processing by applying a long-range dependence modeling algorithm, capturing direct description contents, processing an analysis process by adopting an attention mechanism technology, and ensuring that a complete context is constructed and a key description item set is generated; based on the key description item set, applying a step-by-step denoising generation model algorithm, performing semantic mapping processing, accurately reflecting actual information, adopting an adversarial training technology, automatically matching medical terms closest to the standard, and generating a structured wounded information summary; and based on the structured wounded person information summary, performing evaluation processing on the emergency degree, and generating a wounded person information rapid identification scheme. According to the technical scheme provided by the invention, the decision support capability is improved, and the wounded information identification speed and accuracy are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent information processing technology, and in particular to a method and system for rapid identification of injured person information based on artificial intelligence. Background Art

[0002] With the continuous development of medical emergency response systems, the intelligence and multimodal data processing capabilities of on-site rescue equipment have become particularly important. In emergency rescue scenarios, rescuers need to quickly and accurately identify the information of the injured in order to provide timely treatment and resource allocation. Specifically, through multiple on-site rescue devices (such as smart cameras, sensors, mobile terminals, etc.), multimodal data streams including video, audio, text, etc. are received, and event correlation processing is performed to generate a comprehensive data fragment collection. This process requires the system to have efficient data fusion capabilities and real-time processing capabilities to ensure the comprehensiveness and accuracy of the information.

[0003] At present, most medical emergency response systems rely mainly on traditional data processing methods, such as rule-based text parsing and simple keyword matching. These systems usually first collect data from different devices, and then preliminarily process the data through preset rules or templates. On this basis, some basic natural language processing techniques, such as word frequency statistics and simple semantic analysis, are used to generate a preliminary summary of the casualty information. However, these methods seem to be powerless when faced with complex and changing on-site situations.

[0004] Although the existing solutions can meet the basic needs of emergency response to a certain extent, there are still obvious shortcomings. Traditional methods are difficult to effectively capture and understand long-distance dependencies, resulting in insufficient accuracy in identifying the injured, especially in complex environments, where key information is easily missed; the existing solutions cannot effectively reduce noise when processing structured and unstructured data, so that the generated information summary is often mixed with a large amount of irrelevant information, affecting the accuracy of decision-making; the existing assessment of the condition of the injured mostly relies on manual judgment and lacks the support of automated reasoning engines, making it difficult to quickly and accurately assess the urgency of the injured, thus delaying the best time for treatment. Summary of the invention

[0005] The embodiments of the present application provide a method and system for rapid identification of injured person information based on artificial intelligence, so as to solve the problem of insufficient accuracy in identifying injured person information in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a method for quickly identifying wounded information based on artificial intelligence, comprising:

[0007] Through multiple on-site rescue devices, multimodal data streams are received, event correlation processing is performed, and a comprehensive data segment collection is generated;

[0008] Based on the comprehensive data fragment set, a long-range dependency modeling algorithm is used to perform deep semantic analysis and processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words, ensuring the construction of a complete context and generating a set of key description items;

[0009] Based on the key description item set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate an adversarial environment, automatically match the closest standard medical terms, and generate a structured summary of casualty information;

[0010] Based on the structured casualty information summary, a pre-configured artificial intelligence reasoning engine is activated to evaluate and process the urgency of the casualty's condition and generate a rapid identification plan for the casualty information.

[0011] Optionally, based on the comprehensive data segment set, a long-range dependency modeling algorithm is used to perform deep semantic analysis processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words to ensure the construction of a complete context, a set of key description items is generated, including:

[0012] Based on the set of comprehensive data segments, data fusion is performed through time synchronization and geolocation to construct accident scenarios and generate situational awareness data sets;

[0013] Based on the context-aware dataset, a long-range dependency modeling algorithm is used to parse the text content layer by layer, enhance the ability to understand the long-range dependency relationship in the text, capture the direct description content, and generate preliminary parsing results;

[0014] Based on the preliminary parsing results, the attention mechanism technology is used to process the analysis process, and the importance weights of different words are dynamically adjusted to ensure the construction of a complete context, optimize and refine key description items, and generate a set of intermediate description items;

[0015] Based on the intermediate description item set, integration and optimization processing are performed, and the relationship between the description items is comprehensively analyzed to generate a key description item set.

[0016] Optionally, based on the context-aware dataset, a long-range dependency modeling algorithm is used to parse the text content layer by layer, enhance the ability to understand the long-range dependency relationship in the text, capture the direct description content, and generate preliminary parsing results, including:

[0017] Based on the context-aware data set, preliminary sorting and data cleaning are performed to remove redundant and irrelevant information and generate an optimized text information set;

[0018] Based on the optimized text information set, a long-range dependency modeling algorithm is used to perform layer-by-layer parsing processing, thereby enhancing the ability to understand long-range dependency relationships in the text, capturing direct description content, and generating a detailed semantic parsing report;

[0019] Based on the detailed semantic analysis report, key information extraction is performed to focus on important information points for understanding the accident scene and generate a refined key information set;

[0020] Based on the refined key information set, the information is classified and prioritized according to actual needs to generate preliminary analysis results.

[0021] Optionally, based on the preliminary parsing result, the attention mechanism technology is used to process the analysis process, and by dynamically adjusting the importance weights of different words, a complete context is ensured to be constructed, key description items are optimized and refined, and a set of intermediate description items is generated, including:

[0022] Based on the preliminary analysis results, identifying description items closely related to the accident scene and generating a set of classified description items;

[0023] Based on the set of classified description items, attention mechanism technology is used to perform context optimization processing, and by dynamically adjusting the importance weights of different words, a complete context is ensured to further optimize the accuracy and completeness of the description items, and an optimized context description report is generated;

[0024] Based on the optimized context description report, key description items are screened and refined to ensure the accuracy and pertinence of the information and generate a refined set of key description items;

[0025] Based on the refined key description item set, relationship mapping and context association processing are performed to generate an intermediate description item set.

[0026] Optionally, based on the set of key description items, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate an adversarial environment, automatically match the closest standard medical terms, and generate a structured casualty information summary, including:

[0027] Based on the key description item set, converting into a machine-readable format, extracting key features of each description item, and generating a formatted key feature set;

[0028] Based on the formatted key feature set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, thereby gradually reducing noise to accurately reflect actual information, improve information clarity and accuracy, and generate a denoised semantic mapping result;

[0029] Based on the denoised semantic mapping results, adversarial training technology is used to simulate the adversarial environment, and by simulating the uncertainty and complexity in the real scene, the model performance is optimized, the information point closest to the actual situation is automatically matched, and an optimized semantic mapping report is generated;

[0030] Based on the optimized semantic mapping report, classification, sorting and standardization are performed to effectively organize and present information and generate a structured summary of casualty information.

[0031] Optionally, based on the formatted key feature set, a stepwise denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, and the denoising semantic mapping result is generated by gradually reducing noise to accurately reflect actual information, improve information clarity and accuracy, and include:

[0032] Based on the formatted key feature set, combining structured and unstructured data, performing preliminary data fusion and semantic mapping processing to generate preliminary mapping results;

[0033] Based on the preliminary mapping results, a step-by-step denoising generation model algorithm is used to perform noise reduction processing, removing a portion of the noise after each round of iteration, and gradually reducing the noise component in the data through multiple rounds of iterations to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate a denoised semantic mapping intermediate result;

[0034] Based on the denoised semantic mapping intermediate result, perform detail optimization and correction, correct potential deviations and inconsistencies, and generate an optimized and corrected semantic mapping report;

[0035] Based on the optimized and corrected semantic mapping report, information aggregation and structural transformation are performed to generate a denoised semantic mapping result.

[0036] Optionally, based on the structured casualty information summary, activating a pre-configured artificial intelligence reasoning engine, evaluating and processing the urgency of the casualty condition, and generating a quick identification scheme for the casualty information includes:

[0037] Based on the structured casualty information summary, the key attributes of each casualty are accurately extracted, the time and location of the accident are introduced, and a preprocessing feature set is generated;

[0038] Based on the preprocessed feature set, a preconfigured artificial intelligence reasoning engine is activated to perform a comprehensive evaluation through a deep learning rule base to generate a preliminary urgency score;

[0039] Based on the preliminary urgency score, a weighted calculation is performed on multiple scoring dimensions to determine the final urgency level of each injured person and generate a priority information list of the injured persons;

[0040] Based on the priority information list of the injured, the specific situation and urgency of each injured person are described, and on-site rescue personnel are guided to take appropriate measures to generate a rapid identification plan for the injured information.

[0041] In a second aspect, the embodiment of the present application provides a system for rapid identification of wounded information based on artificial intelligence, including:

[0042] A receiving module, used to receive multimodal data streams through multiple on-site rescue devices, perform event correlation processing, and generate a comprehensive data segment set;

[0043] An analysis module is used to perform deep semantic analysis based on the comprehensive data fragment set using a long-range dependency modeling algorithm, capture direct description content by enhancing the ability to understand long-range dependencies in the text, use attention mechanism technology to process the analysis process, and dynamically adjust the importance weights of different words to ensure the construction of a complete context and generate a set of key description items;

[0044] A processing module is used to perform semantic mapping processing based on the key description item set, using a step-by-step denoising generation model algorithm, combining structured and unstructured data, gradually reducing noise to accurately reflect actual information, using adversarial training technology to simulate an adversarial environment, automatically matching the closest standard medical terms, and generating a structured casualty information summary;

[0045] The activation module is used to activate the pre-configured artificial intelligence reasoning engine based on the structured casualty information summary, evaluate and process the urgency of the casualty condition, and generate a rapid identification plan for the casualty information.

[0046] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an artificial intelligence-based method for rapid identification of casualty information as described in the first aspect.

[0047] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for rapid identification of injured person information based on artificial intelligence as described in the first aspect.

[0048] In an embodiment of the present application, a multimodal data stream is received through multiple on-site rescue devices, event association processing is performed, and a comprehensive data segment set is generated; based on the comprehensive data segment set, a long-range dependency modeling algorithm is used to perform deep semantic analysis processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words, ensuring the construction of a complete context, and generating a set of key description items; based on the key description item set, a step-by-step denoising generation model algorithm is used, combining structured and unstructured data, and semantic mapping processing is performed, by gradually reducing noise to accurately reflect actual information, using adversarial training technology to simulate an adversarial environment, automatically matching the closest standard medical terms, and generating a structured casualty information summary; based on the structured casualty information summary, a pre-configured artificial intelligence reasoning engine is activated, the urgency of the casualty condition is evaluated and processed, and a quick identification plan for casualty information is generated. Multiple on-site rescue devices are used to receive multimodal data streams and perform event association processing to ensure the comprehensiveness and accuracy of the data. The system captures direct description content by enhancing the ability to understand long-distance dependencies in the text, and uses the attention mechanism to dynamically adjust the importance weights of vocabulary to build a complete context, thereby improving the depth and accuracy of semantic parsing. Structured and unstructured data are combined for semantic mapping processing, which gradually reduces noise to accurately reflect actual information, thereby improving the clarity and accuracy of information. Adversarial training technology is used to simulate adversarial environments and automatically match the closest standard medical terms to ensure that the generated information summary complies with medical standards.

[0049] Furthermore, data fusion is performed through time synchronization and geographic positioning, accident scenarios are constructed, and situational awareness datasets are generated, which enhances the authenticity and interpretability of the data. The long-range dependency modeling algorithm is used to parse the text content layer by layer, enhance the ability to understand long-range dependencies in the text, capture direct description content, and improve the accuracy of the preliminary parsing results. Based on the preliminary parsing results, the attention mechanism technology is used to handle the analysis process, dynamically adjust the importance weights of different words, ensure the construction of a complete context, optimize and refine key descriptive items, and improve the quality of the intermediate description item set. Through a comprehensive analysis of the relationship between each descriptive item, a key description item set is generated to ensure that the final output information is highly condensed and representative.

[0050] Furthermore, the set of key descriptive items is converted into a machine-readable format, the key features of each descriptive item are extracted, and a formatted key feature set is generated to facilitate subsequent processing; a step-by-step denoising generative model algorithm is used to combine structured and unstructured data for semantic mapping processing, which gradually reduces noise, accurately reflects actual information, and improves the clarity and accuracy of information; adversarial training technology is used to simulate adversarial environments, and by simulating the uncertainty and complexity in real scenarios, the model performance is optimized, and the information points closest to the actual situation are automatically matched to ensure that the generated semantic mapping report is closer to reality; the optimized semantic mapping report is classified, sorted and standardized to effectively organize and present information, and generate a structured summary of casualty information for quick understanding and application.

[0051] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 A flowchart of a method for rapid identification of injured person information based on artificial intelligence provided in an embodiment of the present application;

[0054] Figure 2 A schematic diagram of the structure of a system for rapid identification of injured person information based on artificial intelligence provided in an embodiment of the present application;

[0055] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0057] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0058] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0059] Figure 1 A flowchart of a method for rapid identification of injured person information based on artificial intelligence is provided for the embodiment of the present application, such as Figure 1 As shown, the method includes:

[0060] 101. Receive multimodal data streams through multiple on-site rescue devices, perform event correlation processing, and generate a comprehensive data segment set;

[0061] In this step, multimodal data stream refers to multiple types of data from different sensors and devices, such as video, audio, text, geographic location information, etc. These data are used to describe the overall situation at the accident scene, provide a rich source of information, and ensure the accuracy and completeness of subsequent processing.

[0062] Event correlation processing is to fuse data from different sources through time synchronization and geolocation technology to ensure the temporal and spatial consistency of the data. It includes aligning the timestamps of data in different modalities and confirming the specific location of the accident based on geographic coordinates.

[0063] The comprehensive data fragment collection is an integration of data fragments that have been preliminarily processed and annotated. It not only contains the original data, but also includes key event fragments that have been preprocessed and associated, providing a solid foundation for subsequent analysis.

[0064] In the embodiment of the present application, first, the system collects multimodal data through multiple on-site rescue devices (such as smart cameras, microphones, mobile terminals, etc.); secondly, the system uses time synchronization and geo-positioning technology to pre-process the collected data to ensure the consistency and accuracy of the data; thirdly, the system performs event correlation processing on the pre-processed data to identify and extract key event fragments; finally, the system integrates all relevant data fragments to generate a complete set of comprehensive data fragments, providing a solid foundation for subsequent steps.

[0065] Suppose at a traffic accident scene, the system first captures the on-site video through the smart camera, the microphone records the audio, and the sensor records the geographic location information; secondly, the system synchronizes this data to the same timeline and confirms the exact location of the accident through geolocation; thirdly, the system identifies the first aid scenes in the video, the cries for help in the audio, and the environmental parameters recorded by the sensor; finally, the system integrates all relevant information to generate a detailed and comprehensive data segment collection to provide complete information support for subsequent analysis.

[0066] 102. Based on the comprehensive data fragment set, use the long-range dependency modeling algorithm to perform deep semantic analysis and processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using the attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words, ensuring the construction of a complete context and generating a set of key description items;

[0067] In this step, the long-range dependency modeling algorithm can capture the relationship between distant words in the text and enhance the understanding of complex sentence structures.

[0068] Deep semantic analysis is a method that uses long-range dependency modeling algorithms to capture the relationships between distant words in text, aiming to enhance the understanding of complex sentence structures.

[0069] The ability to understand long-distance dependencies is used to better capture the logical connections between words that are far apart in a sentence, thereby improving the accuracy of parsing the text content.

[0070] Direct descriptive content refers to specific information clearly stated in the text that is crucial to understanding the accident scene and the condition of the injured.

[0071] The attention mechanism technology dynamically adjusts the importance weights of different words, allowing the model to focus on the most critical information points and build a complete context.

[0072] Importance weights are the relative importance of each word in the current context. By dynamically adjusting these weights, the model can more accurately reflect the true intent of the text.

[0073] The key description item set is the important information points extracted from the text that best reflect the actual situation. It removes redundant information, optimizes and refines the key content describing the condition of the injured, and provides accurate information for subsequent processing.

[0074] In the embodiments of the present application, firstly, the system parses the text content based on a comprehensive data segment set and uses a long-range dependency modeling algorithm to capture the direct description content; secondly, the system adopts the attention mechanism technology to dynamically adjust the importance weights of different words to ensure the construction of a complete context; thirdly, the system optimizes and refines key description items and removes redundant information; finally, the system generates a high-quality set of key description items to provide accurate information for subsequent processing.

[0075] For example, continuing with the above example, suppose the system parses the text content in the comprehensive data fragment collection and identifies the key sentences describing the condition of the injured; secondly, the system uses the attention mechanism technology to adjust the importance weight of each word to highlight the key information; thirdly, the system optimizes and refines the key description items, removes irrelevant words and repeated information; finally, the system generates a set of key description items containing the most important information to ensure that subsequent steps can be processed based on the most accurate data.

[0076] 103. Based on the set of key description items, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate adversarial environments, automatically match the closest standard medical terms, and generate a structured summary of casualty information;

[0077] The step-by-step denoising generative model algorithm is a method that reduces noise and improves information clarity and accuracy through multiple iterations. It combines structured and unstructured data, performs semantic mapping processing, and gradually removes irrelevant noise information to accurately reflect the actual information.

[0078] Adversarial training technology simulates the uncertainty and complexity in the real environment, optimizes model performance, and ensures its robustness in various complex scenarios.

[0079] Standard medical terminology is the recognized standard terminology in the medical field. The process of automatically matching the closest standard medical terminology ensures that the generated information complies with medical standards.

[0080] The formatted key feature set is a set of key description items converted into a machine-readable format for subsequent semantic mapping processing.

[0081] The denoising semantic mapping result is a high-precision semantic representation generated after gradually reducing the noise, which can more accurately reflect the actual situation.

[0082] The optimized semantic mapping report is a further optimization of the denoised semantic mapping results. It improves the adaptability of the model by simulating the uncertainty and complexity in real scenarios.

[0083] The structured casualty information summary is the final standardized information collection, which can accurately reflect the actual information and facilitate medical staff to quickly understand and apply it.

[0084] Progressive noise reduction means gradually reducing the noise in the data through multiple rounds of iterations to ensure that the information finally generated is as pure and accurate as possible.

[0085] Semantic mapping processing refers to converting text content into machine-readable semantic representations for subsequent processing and analysis.

[0086] Adversarial training technology optimizes the performance and accuracy of the model by simulating various situations that may be encountered in the real environment.

[0087] In the embodiments of the present application, firstly, based on a comprehensive data segment set, a long-range dependency modeling algorithm is used to parse the text content and capture the direct description content; secondly, the system adopts the attention mechanism technology to dynamically adjust the importance weights of different words to ensure the construction of a complete context; thirdly, the system optimizes and refines key description items and removes redundant information; finally, the system generates a high-quality set of key description items to provide accurate information for subsequent processing.

[0088] For example, continuing with the above example, suppose the system parses the text content in the comprehensive data fragment collection and identifies the key sentences describing the condition of the injured; secondly, the system uses the attention mechanism technology to adjust the importance weight of each word to highlight the key information; thirdly, the system optimizes and refines the key description items, removes irrelevant words and repeated information; finally, the system generates a set of key description items containing the most important information to ensure that subsequent steps can be processed based on the most accurate data.

[0089] 104. Based on the structured casualty information summary, activate the pre-configured artificial intelligence reasoning engine, evaluate and process the urgency of the casualty condition, and generate a rapid identification plan for the casualty information.

[0090] The pre-configured AI reasoning engine is a specially designed software module that can quickly assess the urgency of a casualty’s condition after receiving a structured casualty information summary.

[0091] The rapid identification plan for the injured is a specific response measure generated based on the assessment results, aimed at guiding rescue personnel to take the most effective rescue actions.

[0092] Assessment and treatment refers to determining the severity of the casualty's condition by analyzing the structured casualty information summary, thereby deciding what emergency measures to take.

[0093] In the embodiment of the present application, first, the pre-configured artificial intelligence reasoning engine is activated to evaluate the received structured summary of the injured person's information; second, the system determines the urgency of the injured person's condition based on the evaluation results; third, the system generates specific response measures and priority rankings; finally, the system outputs a complete rapid identification plan for the injured person's information to guide rescuers to quickly take the most effective rescue actions. In summary, this step ensures the efficiency and accuracy of emergency response.

[0094] For example, continuing with the above example, suppose the system activates a pre-configured artificial intelligence reasoning engine to evaluate the received structured casualty information summary; secondly, the system determines the urgency of the casualty's condition based on the evaluation results and generates a corresponding priority ranking; thirdly, the system generates specific response measures, such as immediately sending to the hospital or providing initial treatment on site; finally, the system outputs a complete rapid identification plan for the casualty information to guide rescue personnel to quickly take the most effective treatment actions.

[0095] In summary, steps 101 to 104 cover the entire process from multimodal data collection to the final generation of a rapid identification scheme for injured person information, aiming to provide an efficient and intelligent medical emergency response system to meet the needs of fast and accurate information processing in emergency rescue scenarios.

[0096] In order to further improve the accuracy of deep semantic analysis, in some embodiments, the deep semantic analysis based on the comprehensive data segment set in step 102 includes: based on the comprehensive data segment set, data fusion is performed through time synchronization and geographic positioning to construct an accident scenario and generate a context-aware data set; based on the context-aware data set, a long-range dependency modeling algorithm is used to parse the text content layer by layer to enhance the ability to understand long-range dependency relationships in the text, capture direct description content, and generate preliminary analysis results; based on the preliminary analysis results, an attention mechanism technology is used to process the analysis process, and the importance weights of different words are dynamically adjusted to ensure the construction of a complete context, optimize and refine key description items, and generate a set of intermediate description items; based on the set of intermediate description items, integration and optimization are performed to comprehensively analyze the relationship between the description items and generate a set of key description items.

[0097] A context-aware dataset refers to a comprehensive data set formed by aligning multimodal data through time and geographic location information. It contains not only the original video, audio, and text data, but also the timestamps and geographic coordinates of these data, ensuring the consistency and accuracy of the data.

[0098] The preliminary parsing result is a preliminary information set generated by parsing the text content layer by layer, capturing the direct description content.

[0099] The intermediate description item set is the intermediate result generated in the process of optimizing and refining key description items. It removes some redundant information but has not yet been fully integrated.

[0100] The key description item set is the important information points extracted from the text that best reflect the actual situation. It removes redundant information, optimizes and refines the key content describing the condition of the injured, and provides accurate information for subsequent processing.

[0101] In the embodiments of the present application, firstly, based on a comprehensive set of data fragments, data fusion is performed through time synchronization and geographic positioning, accident scenarios are constructed, and a situational awareness data set is generated; secondly, the system uses a long-range dependency modeling algorithm to parse the text content layer by layer, enhance the ability to understand long-range dependency relationships in the text, capture direct description content, and generate preliminary parsing results; thirdly, the system uses attention mechanism technology to process the analysis process, and ensures the construction of a complete context by dynamically adjusting the importance weights of different words, optimizing and refining key description items, and generating a set of intermediate description items; finally, the system integrates and optimizes the set of intermediate description items, comprehensively analyzes the relationship between each description item, and generates a set of key description items.

[0102] Here is a specific example:

[0103] Suppose at a traffic accident scene, first, the system uses time synchronization and geolocation technology to fuse data from multiple cameras, microphones, and sensors to build a detailed situational awareness dataset; secondly, the system uses a long-range dependency modeling algorithm to parse the fused text content layer by layer, capture direct descriptions of the accident and the condition of the injured, and generate preliminary parsing results; thirdly, the system uses attention mechanism technology to dynamically adjust the importance weights of different words to ensure the construction of a complete context, optimize and refine key descriptive items, and generate a set of intermediate descriptive items; finally, the system integrates and optimizes the set of intermediate descriptive items, comprehensively analyzes the relationship between each descriptive item, and generates a final set of key descriptive items to ensure that subsequent steps can be processed based on the most accurate data.

[0104] In order to further improve the accuracy and efficiency of text content parsing, in some embodiments, the deep semantic analysis processing based on the context-aware data set described in step 102 includes: based on the context-aware data set, preliminary sorting and data cleaning processing are performed to remove redundant and irrelevant information and generate an optimized text information set; based on the optimized text information set, a long-range dependency modeling algorithm is used to perform layer-by-layer parsing processing to enhance the ability to understand long-range dependency relationships in the text, capture direct description content, and generate a detailed semantic parsing report; based on the detailed semantic parsing report, key information extraction processing is performed to focus on important information points for understanding the accident scene and generate a refined key information set; based on the refined key information set, the information is classified and prioritized according to actual needs to generate a preliminary parsing result.

[0105] In this embodiment, the preliminary arrangement and data cleaning process refers to preprocessing the text information in the context-aware data set, removing redundant and irrelevant information, and generating an optimized text information set.

[0106] Redundant information refers to repeated or unnecessary descriptions, while irrelevant information refers to content that does not contribute to understanding the accident scene.

[0107] The optimized text information set is high-quality text data that has been cleaned to reduce noise and improve the accuracy of subsequent processing.

[0108] The detailed semantic analysis report is a high-precision semantic representation generated after parsing the text content layer by layer. It not only contains the surface meaning of the text, but also deeply explores the potential connections between words.

[0109] Key information extraction processing refers to screening out the most important information points from the detailed semantic analysis report to ensure that the system focuses on the content that is critical to understanding the accident scenario.

[0110] Refining the key information set is to remove redundant information, optimize and refine the key description items, and provide accurate information for subsequent processing.

[0111] Classification and prioritization refers to categorizing information according to actual needs and determining the order of treatment to ensure the rational allocation of resources.

[0112] The preliminary parsing result refers to a set of the most important and relevant information generated after completing layer-by-layer semantic parsing, key information extraction, classification and priority sorting of the optimized text information set.

[0113] In the embodiment of the present application, first, the system performs preliminary sorting and data cleaning based on the context-aware data set, removes redundant and irrelevant information, and generates an optimized text information set; secondly, the system uses a long-range dependency modeling algorithm to perform layer-by-layer analysis to enhance the ability to understand long-range dependency relationships in the text, capture direct description content, and generate a detailed semantic analysis report; thirdly, the system performs key information extraction and processing, focusing on important information points for understanding the accident scene, and generates a refined key information set; finally, the system classifies and prioritizes the information according to actual needs, and generates preliminary analysis results.

[0114] Here is a specific example:

[0115] Suppose at an emergency rescue site of a natural disaster, first, the system performs preliminary organization and data cleaning of the text information in the situational awareness dataset, removes redundant and irrelevant information, and generates an optimized text information set; secondly, the system uses a long-range dependency modeling algorithm to parse the optimized text information layer by layer, enhances the ability to understand long-range dependencies in the text, captures direct description content, and generates a detailed semantic analysis report; thirdly, the system performs key information extraction and processing, focuses on important information points for understanding the accident scene, and generates a refined key information set; finally, the system classifies and prioritizes the information according to actual needs, generates preliminary analysis results, and guides rescue personnel to take the most effective response measures.

[0116] In order to further improve the completeness of context construction and the accuracy of description items, in some embodiments, the deep semantic analysis processing based on the preliminary parsing results in step 102 includes: based on the preliminary parsing results, identifying description items closely related to the accident scene, and generating a set of classified description items; based on the set of classified description items, using attention mechanism technology to perform context optimization processing, by dynamically adjusting the importance weights of different words to ensure the construction of a complete context, further optimize the accuracy and completeness of the description items, and generate an optimized context description report; based on the optimized context description report, perform key description item screening and refinement processing to ensure the accuracy and pertinence of the information and generate a refined set of key description items; based on the refined set of key description items, perform relationship mapping and context association processing to generate a set of intermediate description items.

[0117] In this embodiment, the classified description item set refers to identifying description items closely related to the accident scene from the preliminary parsing results and classifying them according to their content and importance.

[0118] The optimized context description report is a high-precision semantic representation generated after context optimization processing. It not only contains the surface meaning of the text, but also deeply explores the potential connections between words to ensure the accuracy and completeness of the description items.

[0119] The refined set of key descriptive items is a high-quality information set generated by screening and refining the optimized context description report to ensure the accuracy and pertinence of the information.

[0120] Relationship mapping and context association processing refers to mapping the relationships of the refined key description items, ensuring the logical coherence between the description items, and generating a set of intermediate description items for use in subsequent steps.

[0121] The intermediate description item set is a transitional information set generated during the deep semantic analysis process. It is formed after relationship mapping and context association processing based on refined key description items.

[0122] In the embodiment of the present application, firstly, based on the preliminary analysis results, the descriptive items closely related to the accident scene are identified, and a set of classified descriptive items is generated; secondly, the system uses the attention mechanism technology to perform context optimization processing, and ensures the construction of a complete context by dynamically adjusting the importance weights of different words, further optimizes the accuracy and completeness of the descriptive items, and generates an optimized context description report; thirdly, the system screens and refines key descriptive items to ensure the accuracy and pertinence of the information, and generates a refined set of key descriptive items; finally, the system performs relationship mapping and context association processing on the refined key descriptive items, and generates a set of intermediate description items, providing a solid foundation for subsequent steps.

[0123] Here is a specific example:

[0124] Suppose that emergency rescue is needed at the scene of a traffic accident on a highway; first, based on the preliminary analysis results, the system identifies descriptive items closely related to the accident scene, such as "vehicle collision" and "severe bleeding", and generates a set of classified descriptive items; secondly, the system uses attention mechanism technology to optimize the context, dynamically adjusts the importance and relevance of different words, ensures the construction of a complete context, and generates an optimized context description report; thirdly, the system screens and refines the key descriptive items in the optimized context description report to ensure the accuracy and pertinence of the information, and generates a refined set of key descriptive items; finally, the system performs relationship mapping and context association processing on the refined key descriptive items to ensure the logical coherence between the descriptive items, generates a set of intermediate description items, and guides subsequent rescue operations.

[0125] In order to further improve the clarity and accuracy of information, in some embodiments, the semantic mapping processing based on the key description item set in step 103 includes: based on the key description item set, converting into a machine-readable format, extracting the key features of each description item, and generating a formatted key feature set; based on the formatted key feature set, using a step-by-step denoising generation model algorithm, combining structured and unstructured data, performing semantic mapping processing, and gradually reducing noise to accurately reflect actual information, thereby improving information clarity and accuracy and generating a denoised semantic mapping result; based on the denoised semantic mapping result, using adversarial training technology to simulate an adversarial environment, simulating uncertainty and complexity in real scenarios, optimizing model performance, automatically matching information points closest to the actual situation, and generating an optimized semantic mapping report; based on the optimized semantic mapping report, performing classification, sorting and standardization processing, effectively organizing and presenting information, and generating a structured casualty information summary.

[0126] In this embodiment, the formatted key feature set refers to a set of key description items converted into a machine-readable format, which extracts the key features of each description item to ensure that the system can effectively perform subsequent processing.

[0127] The denoising semantic mapping result is a high-precision semantic representation generated after gradually reducing the noise, which can more accurately reflect the actual situation.

[0128] The optimized semantic mapping report is a further optimization of the denoised semantic mapping results. It improves the adaptability of the model by simulating the uncertainty and complexity in real scenarios and automatically matches the information points that are closest to the actual situation.

[0129] The structured casualty information summary is the final standardized information collection. It is classified, sorted and standardized to effectively organize and present information, making it easy for medical staff to quickly understand and apply it.

[0130] In the embodiment of the present application, first, the system converts a set of key descriptive items into a machine-readable format, extracts the key features of each descriptive item, and generates a formatted key feature set; secondly, the system uses a step-by-step denoising generation model algorithm, combines structured and unstructured data, and performs semantic mapping processing, thereby gradually reducing noise to accurately reflect actual information, improve information clarity and accuracy, and generate denoised semantic mapping results; thirdly, the system uses adversarial training technology to simulate an adversarial environment, optimizes model performance by simulating the uncertainty and complexity in real scenarios, automatically matches information points that are closest to the actual situation, and generates an optimized semantic mapping report; finally, the system classifies, sorts, and standardizes the optimized semantic mapping report, effectively organizes and presents information, and generates a structured casualty information summary.

[0131] Here is a specific example:

[0132] Assuming that in a high-rise building fire rescue scenario, first, the system collects multimodal data from smart devices worn by firefighters, on-site cameras, and emergency call records; secondly, the system converts these key descriptive items into a machine-readable format, extracts the key features of each descriptive item, and generates a formatted key feature set; thirdly, the system uses a step-by-step denoising generative model algorithm, combines structured and unstructured data, performs semantic mapping processing, gradually reduces noise, improves information clarity and accuracy, and generates denoised semantic mapping results; finally, the system uses adversarial training technology to simulate the uncertainty and complexity in the real environment, optimizes model performance, automatically matches the information points closest to the actual situation, generates an optimized semantic mapping report, and classifies and sorts the report and standardizes it to generate a structured summary of the casualty information, guiding rescue personnel to quickly determine the priority rescue objects and action plans.

[0133] Optionally, based on the formatted key feature set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, and the actual information is accurately reflected by gradually reducing the noise, thereby improving the clarity and accuracy of the information and generating a denoised semantic mapping result, including: based on the formatted key feature set, combined with structured and unstructured data, preliminary data fusion and semantic mapping processing are performed to generate a preliminary mapping result; based on the preliminary mapping result, a step-by-step denoising generation model algorithm is used to perform noise reduction processing, and a part of the noise is removed after each round of iteration, and the noise component in the data is gradually reduced through multiple rounds of iterations to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate a denoised semantic mapping intermediate result; based on the denoised semantic mapping intermediate result, detail optimization and correction are performed to correct potential deviations and inconsistencies, and an optimized and corrected semantic mapping report is generated; based on the optimized and corrected semantic mapping report, information aggregation and structured conversion are performed to generate a denoised semantic mapping result.

[0134] In this embodiment, the preliminary mapping result refers to the initial semantic representation generated after combining structured and unstructured data. It preliminarily reflects the core meaning of the text content through preliminary data fusion and semantic mapping processing.

[0135] The denoising semantic mapping intermediate result is an intermediate state generated in the process of gradually reducing noise, which gradually improves the clarity and accuracy of the information.

[0136] Detail optimization and correction refers to correcting the intermediate results of semantic mapping after denoising, resolving potential deviations and inconsistencies, and ensuring the integrity and consistency of information.

[0137] The optimized and corrected semantic mapping report is a high-precision semantic representation after detail optimization and correction, which can more accurately reflect the actual situation.

[0138] Information aggregation and structured conversion refers to integrating the optimized semantic mapping reports and converting them into a structured form to generate the final denoised semantic mapping results.

[0139] The denoised semantic mapping result is the final output after multiple rounds of processing and optimization. It accurately reflects the actual information and significantly improves the clarity and accuracy of the information.

[0140] In the embodiment of the present application, first, the system performs preliminary data fusion and semantic mapping processing based on the formatted key feature set and combines structured and unstructured data to generate preliminary mapping results; secondly, the system uses a step-by-step denoising generation model algorithm to perform noise reduction processing, removes part of the noise after each round of iteration, and gradually reduces the noise component in the data through multiple rounds of iterations to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate a denoised semantic mapping intermediate result; thirdly, the system optimizes and corrects the details of the denoised semantic mapping intermediate result, corrects potential deviations and inconsistencies, and generates an optimized and corrected semantic mapping report; finally, the system performs information aggregation and structured conversion on the optimized and corrected semantic mapping report to generate a denoised semantic mapping result.

[0141] Here is a specific example:

[0142] Assuming that at the scene of an explosion at a chemical plant, first, the system performs preliminary data fusion and semantic mapping based on the formatted key feature set, combined with structured and unstructured data collected from multiple sensors and eyewitness descriptions, to generate preliminary mapping results; secondly, the system uses a step-by-step denoising generative model algorithm to perform noise reduction processing, removing part of the noise after each round of iteration, and gradually reducing the noise component in the data through multiple rounds of iterations to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate a denoised semantic mapping intermediate result; thirdly, the system optimizes and corrects the details of the denoised semantic mapping intermediate result, corrects potential deviations and inconsistencies, and generates an optimized and corrected semantic mapping report; finally, the system performs information aggregation and structured conversion on the optimized and corrected semantic mapping report to generate a denoised semantic mapping result, guiding rescue personnel to quickly take the most effective response measures.

[0143] In order to further improve the accuracy and response speed of the emergency assessment of the injured person's condition, in some embodiments, the artificial intelligence reasoning processing based on the structured injured person information summary described in step 104 includes: based on the structured injured person information summary, accurately extracting the key attributes of each injured person, introducing the time and place of the accident, and generating a preprocessing feature set; based on the preprocessing feature set, activating the pre-configured artificial intelligence reasoning engine, performing a comprehensive assessment through a deep learning rule base, and generating a preliminary urgency score; based on the preliminary urgency score, performing weighted calculations on multiple scoring dimensions, determining the final urgency level of each injured person, and generating a list of priority information on the injured; based on the list of priority information on the injured, describing the specific situation and urgency of each injured person, guiding on-site rescue personnel to take appropriate measures, and generating a rapid identification plan for injured information.

[0144] In this embodiment, the preprocessing feature set refers to a data set generated by accurately extracting key attributes of each injured person from the structured injured person information summary and introducing contextual information such as the time and place of the accident.

[0145] The deep learning rule base is a series of trained and optimized algorithm models that are used to comprehensively evaluate the input data and generate a preliminary urgency score.

[0146] The preliminary urgency score is an initial score generated based on a set of preprocessed features and evaluated by a deep learning rule base. It reflects the urgency of the casualty's current condition.

[0147] Weighted calculation refers to assigning weights according to the importance of different scoring dimensions and determining the final urgency level after comprehensively considering multiple factors.

[0148] The priority information list of the wounded is a list of all the wounded sorted according to their urgency, ensuring that rescue resources can be used first for the wounded who need them most.

[0149] The rapid identification plan for the injured is to describe in detail the specific situation and urgency of each injured person based on the priority information list of the injured, and guide on-site rescue personnel to take appropriate treatment measures.

[0150] In the embodiment of the present application, first, based on the structured casualty information summary, the key attributes of each casualty are accurately extracted, the time and place of the accident are introduced, and a preprocessing feature set is generated; secondly, the system activates the pre-configured artificial intelligence reasoning engine, and comprehensively evaluates the preprocessing feature set through the deep learning rule base to generate a preliminary urgency score; thirdly, the system performs weighted calculations on multiple scoring dimensions, determines the final urgency level of each casualty, and generates a list of casualty priority information; finally, the system describes the specific situation and urgency of each casualty based on the casualty priority information list, guides on-site rescue personnel to take appropriate measures, and generates a rapid identification plan for casualty information.

[0151] Here is a specific example:

[0152] Assuming that at the scene of a large-scale traffic accident, first, the system accurately extracts the key attributes of each injured person based on the structured casualty information summary, introduces the time and place of the accident, and generates a pre-processed feature set; secondly, the system activates the pre-configured artificial intelligence reasoning engine, comprehensively evaluates the pre-processed feature set through the deep learning rule base, and generates a preliminary urgency score; thirdly, the system performs weighted calculations on multiple scoring dimensions, determines the final urgency level of each injured person, and generates a list of priority information on the injured; finally, the system describes the specific situation and urgency of each injured person based on the priority information list of the injured, guides on-site rescue personnel to take appropriate measures, and generates a rapid identification plan for the injured information to ensure that the most critically ill injured receive priority treatment.

[0153] This application considers that in order to solve the problem in the prior art that the key information and emotional elements cannot be accurately captured in complex text analysis due to insufficient understanding of long-distance dependencies in the text, the invention embodiment proposes this optional solution. To solve the above problem, a new optional solution is proposed, which includes:

[0154] Based on the optimized text information set, the long-range dependency modeling algorithm is used to perform layer-by-layer parsing processing, enhance the ability to understand long-range dependency relationships in the text, capture direct description content, and generate a detailed semantic parsing report, including:

[0155] Based on the optimized text information set, syntactic analysis and dependency parsing are performed to convert the text information into a high-dimensional vector representation to capture the semantic similarity between words and generate a long-distance dependency score;

[0156] The long-distance dependency score is calculated using the following formula:

[0157]

[0158] Among them, LDR iis the long-distance dependency score of the lth text segment; w j is the importance weight of the Jth word; β is the Gaussian kernel parameter; Dist(i,j) 2 is the distance between text segment i and word j; j is the word index in the text segment, ranging from 1 to N; N is the total number of words in the text segment; α is the periodic influence weight; ω is the periodic frequency parameter; Freq(i,j) is the frequency of word j in text segment i; η is the contextual influence weight; v k is the importance weight of the k-th context feature; Context(i,k) is the k-th context feature of text segment i; k is the index of the considered context feature type, from 1 to K; K is the number of considered context feature types;

[0159] Based on the long-distance dependency score, the complex emotional elements implied in the text are identified through the sentiment dictionary, the influence of contextual features is considered, the extraction is mapped to the corresponding vector space, and a nonlinear activation function is introduced to adjust the contribution of each part to generate a comprehensive semantic analysis score;

[0160] The comprehensive semantic parsing score is calculated using the following formula:

[0161]

[0162] Among them, SemanticScore i is the comprehensive semantic parsing score of the i-th text segment; LDR i is the long-distance dependency score of the i-th text segment; α' is a balance parameter used to adjust the importance of long-distance dependency and sentiment elements; e k is the intensity of the kth implicit complex emotional element; γ is the activation function parameter; Emotion(i,k) is the correlation between text segment i and the kth implicit complex emotional element; k is the index of the type of implicit complex emotional element considered, from 1 to M; M is the number of types of implicit complex emotional elements considered; δ is the context influence weight; h l is the importance weight of the k-th context feature; ω is the periodic frequency parameter; Context(i,l) is the l-th context feature of text segment i; l is the index of the considered context feature type, from 1 to P; P is the number of considered context feature types;

[0163] Based on the comprehensive semantic analysis score, multi-dimensional information integration and visualization processing are performed to present key information for quick reference, and a concise summary is automatically generated through natural language generation technology to generate a detailed semantic analysis report.

[0164] This method aims to parse text content more accurately by evaluating the long-distance dependencies between text fragments and the emotional elements they contain through quantitative methods. This not only helps to improve the understanding of text structure, but also better captures the emotional information in the text, allowing the system to parse the text more comprehensively and meticulously, especially for scenarios such as medical emergency response that require fast and accurate information processing.

[0165] In the long-range dependency score, the vocabulary weight term w j ·exp(-β·Dist(i,j) 2 0: This item is used to measure the importance of each word and its distance from the text segment. By introducing the Gaussian kernel function exp(-β·Dist(i,j) 2 ), which can effectively capture the dependencies between distant words and ensure that the system can identify long-distance semantic associations; Periodic frequency term 1+α·sin(ω·Freq(i,j)): This sub-item adjusts the effect of word frequency on dependencies, and simulates the repetitive and periodic features in human language by introducing periodic fluctuations sin(ω·Freq(i,j)), enhancing the model’s understanding of text structure; Context influence term This sub-item considers the impact of context features on dependencies by introducing context features Context(i,k) and their importance weights v k ,Ensure that the model can capture the semantic changes of text snippets in different contexts;

[0166] Among them, the importance weight of the word w j It can be calculated by term frequency (TF) and inverse document frequency (IDF). The specific formula is w j =TF-IDF(j); Gaussian kernel parameter β is usually determined by cross-validation to find the optimal distance attenuation effect; the distance between text segments Dist(i,j) is calculated based on the position of the vocabulary, such as the character distance between words or the relative position in the sentence; the periodic influence weight α and the periodic frequency parameter ω can be optimized according to the experimental data and adjusted to the optimal value through multiple experiments; the vocabulary frequency Freq(i,j) is obtained by statistical methods, that is, the number of times vocabulary j appears in the text segment |; the contextual influence weight η is obtained by training the model and is usually optimized in the model training stage; the importance weight v of the contextual feature k It is also obtained through training models to ensure that the importance of context features is reasonably distributed; the context feature Context(i,k) is extracted through natural language processing tools, such as part-of-speech tagging, dependency syntax analysis, etc.

[0167] In the comprehensive semantic parsing score, the long-distance dependency contribution term α'·LDR i: This item is used to balance the impact of long-distance dependency scores on the comprehensive semantic analysis score, ensuring that the system can fully consider the structure and content of the text; sentiment intensity item This item adjusts the contribution of emotional elements by introducing a nonlinear activation function, tanh, to ensure that the system can accurately capture the implicit emotions in the text, and balance the influence of different emotional elements through weighted averaging; contextual influence item This sub-item further considers the impact of contextual features on semantic parsing. It simulates the changes of contextual features by introducing periodic fluctuations sin(ω·Context(i,l)) to ensure that the system can capture the semantic changes of text fragments in different contexts.

[0168] Among them, the balance parameter α' is adjusted through experiments to ensure the best balance between long-distance dependencies and emotional elements; the long-distance dependency score LDR i : Calculated by the first formula; the intensity of the complex emotional elements implied e k It is obtained through sentiment dictionary matching and sentiment classification algorithm; the activation function parameter γ is optimized through experiments to ensure the best effect of the activation function; the correlation between the text fragment and the sentiment element Emotion(i,k) is calculated through the sentiment dictionary and sentiment classification algorithm; the context influence weight δ is obtained through the training model to ensure that the importance of context features is reasonably distributed; the importance weight of context features h l It is also obtained through training models to ensure that the importance of context features is reasonably distributed; the context feature Context(i,l) is extracted through natural language processing tools, such as part-of-speech tagging, dependency syntax analysis, etc.; the periodic frequency parameter ω is optimized according to experimental data to ensure the best effect of periodic fluctuations;

[0169] Suppose that in a hospital emergency room scenario, the system receives an emergency report describing the conditions of multiple patients. In order to quickly assess the urgency of each patient and guide treatment, the system needs to deeply analyze the text;

[0170] Assume that w j =0.4; β=0.7; Dist(i,j)=3; α=0.9; ω=2π; Freq(i,j)=5; η=0.6; v k =0.8;Context(i,k)=1.2;

[0171]

[0172] Assume α' = 0.7; LDR i =0.82; e k =0.6; γ = 0.4; Emotion(i,k) = 0.8; δ = 0.5; hl =0.7; Context(i,l)=1.5; ω=2π;

[0173]

[0174] Assuming that the threshold is set to 0.85, since the calculated result 0.91 is greater than the set threshold, it indicates that the text fragment contains important long-distance dependencies and strong emotional elements, and needs to be processed first. This is because a higher comprehensive semantic parsing score reflects the importance and urgency of the text fragment, ensuring that the system can quickly identify and process key information, thereby improving the efficiency and accuracy of the rescue operation. Through the above steps, the accurate parsing of the information of the wounded and sick is ensured, the reliability and scientificity of the rescue operation are improved, and the accuracy and response speed of the entire emergency response system are enhanced.

[0175] This application considers that in order to solve the technical problem that the existing technology cannot accurately capture key information and emotional elements in complex text analysis due to insufficient processing of noise components in the text, the invention embodiment proposes this optional solution. In order to solve the above problem, a new optional solution is proposed, which includes:

[0176] Based on the preliminary mapping results, a step-by-step denoising generation model algorithm is used to perform noise reduction processing. After each round of iteration, a portion of the noise is removed. Through multiple rounds of iterations, the noise component in the data is gradually reduced to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate the denoising semantic mapping intermediate results, including:

[0177] Based on the preliminary mapping results, the importance and uniqueness of each word in the text are quantified through word frequency-inverse document frequency analysis, the context feature score is used to evaluate the context relevance of the word, and sentiment analysis is introduced to identify potential sentiment elements to generate denoising probability;

[0178] The denoising probability is calculated using the following formula:

[0179]

[0180] Among them, NP t (i) is the denoising probability of text segment i after the tth iteration; t is the iteration round; i is the text segment index; w j is the importance weight of the jth word; λ is the adjustment parameter used to adjust the impact of the TF-IDF value; TF(i,j) is the word frequency of word j in text segment i; T is the total number of documents; DF(i,j) is the number of documents containing word j; j is the word index in the text segment, from 1 to N; N is the total number of words in the text segment; ∈ is the context influence weight; v kis the importance weight of the k-th context feature; CS(i,k) is the score of the k-th context feature of text segment i; k is the index of the considered context feature type, from 1 to K; K is the number of considered context feature types;

[0181] Based on the noise probability, combined with the influence of emotional elements, the emotional intensity weight is adjusted through a nonlinear activation function, the role of position deviation characteristics is considered, and a periodic function is used to capture the distribution pattern of words in the text to generate a cleanliness score;

[0182] The cleanliness score is calculated using the following formula:

[0183]

[0184] Among them, CP t+1 (i) is the cleanliness score of text segment i after the t+1th iteration; NP t (i) is the noise probability of text segment i after the tth iteration; m is the index of the type of sentiment element considered, from 1 to M; M is the number of types of sentiment elements considered; σ m is the emotional impact weight; m is the activation function parameter; ES(i,m) is the intensity of the mth implicit complex emotional element in text segment i; δ' is the position deviation weight; h q is the importance weight of the qth position deviation feature; ω q is the periodic frequency parameter; PB(i,q) is the qth position deviation feature of text segment i; q is the index of the considered position deviation feature type, from 1 to L; L is the number of considered position deviation feature types; ζ is the correlation weight; u n is the importance weight of the nth relevance feature; RS(i,n) is the score of the nth relevance feature of text segment i; n is the index of the type of relevance feature considered, from 1 to Q; Q is the number of types of relevance features considered; η' and μ are confidence adjustment parameters; CF(i) is the confidence score of the model for text segment i;

[0185] Based on the cleanliness score, high-scoring paragraphs are processed first, low-scoring partial noise is gradually removed, and vocabulary importance weights are continuously updated to ensure that each iteration more accurately reflects actual information for recombination to generate denoised semantic mapping intermediate results.

[0186] This method aims to parse the text content more accurately by quantitatively evaluating the noise components and emotional elements in the text fragments. This not only helps to improve the understanding of the text structure, but also better captures the emotional information in the text, allowing the system to parse the text more comprehensively and meticulously, especially for scenarios such as medical emergency response that require fast and accurate information processing.

[0187] In the denoising probability, the vocabulary importance term w j ·(1-exp(-λ·TF(i,j)·log(T / (DF(i,j))))): This item is used to measure the importance and uniqueness of each word in the text. The exponential decay function exp(-λ·TF(i,j)·log(T / (DF(i,j)))) is introduced to adjust the importance weight of the word to ensure that high-frequency but ubiquitous words do not excessively affect the results; context influence item This sub-item considers the impact of contextual features on the relevance of lexical contexts by introducing the contextual feature score CS(i,k) and its importance weight v k ,Ensure that the model can capture the semantic changes of words in different contexts and enhance the understanding of text structure;

[0188] Among them, the importance weight of the word w j Through experimental adjustment, ensure that the importance of vocabulary is reasonably distributed; the adjustment parameter λ is determined through cross-validation to find the optimal TF-IDF value adjustment effect; the total number of documents T is obtained by counting the number of all documents; the number of documents containing vocabulary j DF(i,j) is obtained by counting the number of documents containing vocabulary j; the context influence weight ∈ is obtained by training the model to ensure that the importance of context features is reasonably distributed; the importance weight v of context features k It is also obtained through training models; the context feature score CS(i,k) is extracted through natural language processing tools, such as part-of-speech tagging, dependency syntax analysis, etc.

[0189] In the cleanliness rating, the noise probability term NP t (i): reflects the noise probability after the previous iteration, ensuring the accumulation of denoising effect after each iteration; emotional impact item By introducing the nonlinear activation function tanh to adjust the emotional intensity weight, we can ensure that the system can accurately capture the implicit emotions in the text and balance the influence of different emotional elements through weighted averaging; position deviation term Use the periodic function sin to capture the distribution pattern of words in the text, simulate the repetitive and periodic features in human language, and enhance the model's understanding of the text structure; correlation feature items This sub-item considers the role of position deviation characteristics by introducing the correlation feature RS(i,n) and its importance weight u n , ensuring that the model can capture the semantic changes of words at different positions; Confidence adjustment term 1+η'·exp(-μ·CF(i)): By introducing the confidence score CF(i), the model's confidence in text segment i is reasonably adjusted to avoid overfitting or underfitting problems;

[0190] Among them, the noise probability NP t (i) Calculated by the first formula; emotional influence weight σ m Obtained through sentiment dictionary matching and sentiment classification algorithm; activation function parameter γ m Through experimental optimization, the activation function is ensured to be optimal; the emotional element strength ES(i,m) is calculated through the emotional dictionary and emotional classification algorithm; the position deviation weight δ' is adjusted through experiments to ensure that the importance of the position deviation feature is reasonably distributed; the importance weight h of the position deviation feature q Obtained by training the model; period frequency parameter ω q Optimize according to experimental data to ensure the best effect of periodic fluctuations; extract the position deviation feature PB(i,q) through natural language processing tools, such as the position information of words; adjust the relevance weight ζ through experiments to ensure that the importance of the relevance feature is reasonably distributed; the importance weight u of the relevance feature n Obtained through training model; correlation feature score RS(i,n) is extracted through natural language processing tools; confidence adjustment parameters η' and μ are adjusted through experiments to ensure the best confidence adjustment effect; confidence score CF(i) is obtained through model prediction;

[0191] Suppose that in a hospital emergency room scenario, the system receives an emergency report describing the conditions of multiple patients. In order to quickly assess the urgency of each patient and guide treatment, the system needs to deeply analyze the text;

[0192] Assume that w j =0.5; λ=0.8; TF(i,j)=3; T=1000; DF(i,j)=20; ∈=0.7; v k =0.6; CS(i,k)=1.5;

[0193]

[0194] Assume NP t (i) = 0.78; σ m =0.4;γ m =0.5; ES(i,m)=0.9; δ'=0.6; h q =0.7;ω q =2π; PB(i,q)=1.2; ζ=0.8; u n =0.5; RS(i,n)=1.1; η'=0.9; μ=0.6; CF(i)=0.8;

[0195]

[0196] Assuming that the threshold is set to 0.85, since the calculated result 0.89 is greater than the set threshold, it shows that the text segment has a higher cleanliness score after multiple rounds of iterative denoising, reflecting the importance and accuracy of the text segment. This is because a higher cleanliness score means that the noise in the text has been effectively removed, retaining key information, ensuring that the system can quickly identify and process key information, thereby improving the efficiency and accuracy of the rescue operation. Through the above steps, the accurate analysis of the information of the wounded and sick is ensured, the reliability and scientificity of the rescue operation are improved, and the accuracy and response speed of the entire emergency response system are enhanced.

[0197] Figure 2 A schematic diagram of a system for rapid identification of injured persons based on artificial intelligence is provided for the present application embodiment. Figure 2 As shown, the device comprises:

[0198] The receiving module 21 is used to receive multimodal data streams through multiple on-site rescue devices, perform event correlation processing, and generate a comprehensive data segment set;

[0199] An analysis module 22 is used to perform deep semantic analysis based on the comprehensive data segment set using a long-range dependency modeling algorithm, capture direct description content by enhancing the ability to understand long-range dependencies in the text, use attention mechanism technology to process the analysis process, and dynamically adjust the importance weights of different words to ensure the construction of a complete context and generate a set of key description items;

[0200] The processing module 23 is used to perform semantic mapping processing based on the key description item set by using a step-by-step denoising generation model algorithm, combining structured and unstructured data, gradually reducing noise to accurately reflect actual information, using adversarial training technology to simulate an adversarial environment, automatically matching the closest standard medical terms, and generating a structured casualty information summary;

[0201] The activation module 24 is used to activate the pre-configured artificial intelligence reasoning engine based on the structured casualty information summary, evaluate the urgency of the casualty condition, and generate a rapid identification plan for the casualty information.

[0202] Figure 2 The artificial intelligence-based rapid identification system for injured persons can be implemented Figure 1 The implementation principle and technical effect of the method for rapid identification of wounded information based on artificial intelligence described in the embodiment shown are not repeated here. The specific way in which each module and unit performs operations in the system for rapid identification of wounded information based on artificial intelligence in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0203] In one possible design, Figure 2 The artificial intelligence-based rapid identification system for injured persons in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0204] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0205] The processing component 32 is used to: receive multimodal data streams through multiple on-site rescue devices, perform event association processing, and generate a comprehensive data segment set; based on the comprehensive data segment set, use a long-range dependency modeling algorithm to perform deep semantic analysis processing, capture direct description content by enhancing the ability to understand long-range dependencies in the text, use attention mechanism technology to process the analysis process, and dynamically adjust the importance weights of different words to ensure the construction of a complete context and generate a set of key description items; based on the key description item set, use a step-by-step denoising generation model algorithm, combine structured and unstructured data, perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate an adversarial environment, automatically match the closest standard medical terms, and generate a structured casualty information summary; based on the structured casualty information summary, activate a pre-configured artificial intelligence reasoning engine, evaluate the urgency of the casualty condition, and generate a quick identification plan for the casualty information.

[0206] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0207] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0208] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0209] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0210] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0211] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0212] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a method for quickly identifying wounded person information based on artificial intelligence.

[0213] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0214] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0215] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0216] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for rapid identification of wounded information based on artificial intelligence, characterized in that: include: Receive multimodal data streams through multiple on-site rescue devices, perform event correlation processing, and generate a comprehensive data segment collection; Based on the comprehensive data fragment set, a long-range dependency modeling algorithm is used to perform deep semantic analysis and processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words, ensuring the construction of a complete context and generating a set of key description items; Based on the key description item set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate an adversarial environment, automatically match the closest standard medical terms, and generate a structured summary of casualty information; Based on the structured casualty information summary, a pre-configured artificial intelligence reasoning engine is activated to evaluate and process the urgency of the casualty's condition and generate a rapid identification plan for the casualty information.

2. The method according to claim 1, characterized in that Based on the comprehensive data fragment set, the long-range dependency modeling algorithm is used to perform deep semantic analysis and processing, by enhancing the ability to understand long-range dependencies in the text, capturing direct description content, using the attention mechanism technology to process the analysis process, and by dynamically adjusting the importance weights of different words, ensuring the construction of a complete context, and generating a set of key description items, including: Based on the set of comprehensive data segments, data fusion is performed through time synchronization and geolocation to construct accident scenarios and generate situational awareness data sets; Based on the context-aware dataset, a long-range dependency modeling algorithm is used to parse the text content layer by layer, enhance the ability to understand the long-range dependency relationship in the text, capture the direct description content, and generate preliminary parsing results; Based on the preliminary parsing results, the attention mechanism technology is used to process the analysis process, and the importance weights of different words are dynamically adjusted to ensure the construction of a complete context, optimize and refine key description items, and generate a set of intermediate description items; Based on the intermediate description item set, integration and optimization processing are performed, and the relationship between the description items is comprehensively analyzed to generate a key description item set.

3. The method according to claim 2, characterized in that Based on the context-aware dataset, the long-range dependency modeling algorithm is used to parse the text content layer by layer, enhance the ability to understand the long-range dependency relationship in the text, capture the direct description content, and generate preliminary parsing results, including: Based on the context-aware data set, preliminary sorting and data cleaning are performed to remove redundant and irrelevant information and generate an optimized text information set; Based on the optimized text information set, a long-range dependency modeling algorithm is used to perform layer-by-layer parsing processing, thereby enhancing the ability to understand long-range dependency relationships in the text, capturing direct description content, and generating a detailed semantic parsing report; Based on the detailed semantic analysis report, key information extraction is performed to focus on important information points for understanding the accident scene and generate a refined key information set; Based on the refined key information set, the information is classified and prioritized according to actual needs to generate preliminary analysis results.

4. The method according to claim 2, characterized in that: Based on the preliminary parsing results, the attention mechanism technology is used to process the analysis process, and the importance weights of different words are dynamically adjusted to ensure the construction of a complete context, optimize and refine key description items, and generate an intermediate description item set, including: Based on the preliminary analysis results, identifying description items closely related to the accident scene and generating a set of classified description items; Based on the set of classified description items, attention mechanism technology is used to perform context optimization processing, and by dynamically adjusting the importance weights of different words, a complete context is ensured to further optimize the accuracy and completeness of the description items, and an optimized context description report is generated; Based on the optimized context description report, key description items are screened and refined to ensure the accuracy and pertinence of the information and generate a refined set of key description items; Based on the refined key description item set, relationship mapping and context association processing are performed to generate an intermediate description item set.

5. The method according to claim 1, characterized in that Based on the key description item set, the step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, gradually reduce noise to accurately reflect actual information, use adversarial training technology to simulate adversarial environment, automatically match the closest standard medical terms, and generate a structured casualty information summary, including: Based on the key description item set, converting into a machine-readable format, extracting key features of each description item, and generating a formatted key feature set; Based on the formatted key feature set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, thereby gradually reducing noise to accurately reflect actual information, improve information clarity and accuracy, and generate a denoised semantic mapping result; Based on the denoised semantic mapping results, adversarial training technology is used to simulate the adversarial environment, and by simulating the uncertainty and complexity in the real scene, the model performance is optimized, the information point closest to the actual situation is automatically matched, and an optimized semantic mapping report is generated; Based on the optimized semantic mapping report, classification, sorting and standardization are performed to effectively organize and present information and generate a structured summary of casualty information.

6. The method according to claim 5, characterized in that Based on the formatted key feature set, a step-by-step denoising generation model algorithm is used to combine structured and unstructured data to perform semantic mapping processing, thereby gradually reducing noise to accurately reflect actual information, improve information clarity and accuracy, and generate denoising semantic mapping results, including: Based on the formatted key feature set, combining structured and unstructured data, performing preliminary data fusion and semantic mapping processing to generate preliminary mapping results; Based on the preliminary mapping results, a step-by-step denoising generation model algorithm is used to perform noise reduction processing, removing a portion of the noise after each round of iteration, and gradually reducing the noise component in the data through multiple rounds of iterations to accurately reflect the actual information, improve the clarity and accuracy of the information, and generate a denoised semantic mapping intermediate result; Based on the denoised semantic mapping intermediate result, perform detail optimization and correction, correct potential deviations and inconsistencies, and generate an optimized and corrected semantic mapping report; Based on the optimized and corrected semantic mapping report, information aggregation and structural transformation are performed to generate a denoised semantic mapping result.

7. The method according to claim 1, characterized in that Based on the structured casualty information summary, the pre-configured artificial intelligence reasoning engine is activated to evaluate and process the urgency of the casualty condition and generate a quick identification scheme for the casualty information, including: Based on the structured casualty information summary, the key attributes of each casualty are accurately extracted, the time and location of the accident are introduced, and a preprocessing feature set is generated; Based on the preprocessed feature set, a preconfigured artificial intelligence reasoning engine is activated to perform a comprehensive evaluation through a deep learning rule base to generate a preliminary urgency score; Based on the preliminary urgency score, a weighted calculation is performed on multiple scoring dimensions to determine the final urgency level of each injured person and generate a priority information list of the injured persons; Based on the priority information list of the injured, the specific situation and urgency of each injured person are described, and on-site rescue personnel are guided to take appropriate measures to generate a rapid identification plan for the injured information.

8. A system for rapid identification of casualty information based on artificial intelligence, characterized in that: include: A receiving module, used to receive multimodal data streams through multiple on-site rescue devices, perform event correlation processing, and generate a comprehensive data segment set; An analysis module is used to perform deep semantic analysis based on the comprehensive data fragment set using a long-range dependency modeling algorithm, capture direct description content by enhancing the ability to understand long-range dependencies in the text, use attention mechanism technology to process the analysis process, and dynamically adjust the importance weights of different words to ensure the construction of a complete context and generate a set of key description items; A processing module is used to perform semantic mapping processing based on the key description item set, using a step-by-step denoising generation model algorithm, combining structured and unstructured data, gradually reducing noise to accurately reflect actual information, using adversarial training technology to simulate an adversarial environment, automatically matching the closest standard medical terms, and generating a structured casualty information summary; The activation module is used to activate the pre-configured artificial intelligence reasoning engine based on the structured casualty information summary, evaluate and process the urgency of the casualty condition, and generate a rapid identification plan for the casualty information.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an artificial intelligence-based method for rapid identification of injured person information as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, an artificial intelligence-based method for quickly identifying wounded information as described in any one of claims 1 to 7 is implemented.

Images