Fire alarm processing method and device, computer program product and storage medium
By combining a large language model with a fire safety knowledge base, fire alarm information is processed automatically, solving the problems of omissions and biases in manual summarization and achieving efficient and accurate alarm summarization and command and dispatch support.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DACE INFORMATION TECH CO LTD
- Filing Date
- 2025-12-03
- Publication Date
- 2026-04-24
AI Technical Summary
Existing methods for handling fire emergencies rely on manual summarization, which is prone to omissions and errors, resulting in insufficient accuracy in emergency judgment and command and dispatch.
The system uses a large language model combined with a fire safety knowledge base to analyze and integrate various original emergency information, extract key tags, and generate an emergency summary report, including information correction and contextual reasoning to complete the report and remove unusable information.
It has enabled automated processing and precise summarization of alarm information, improved the accuracy of alarm judgment and command and dispatch, and enhanced the scientific nature and efficiency of fire command decision-making.
Smart Images
Figure CN121920807A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of fire alarm handling technology, and further to a fire alarm handling method and apparatus, computer program product, and storage medium. Background Technology
[0002] With the accelerating pace of urbanization, public safety issues are becoming increasingly prominent, among which fire fighting and emergency rescue tasks are particularly complex and demanding. In recent years, the number of emergency calls received by fire departments at all levels in my country has been rising year by year, and the types of incidents are becoming more diverse, covering various scenarios such as fires, traffic accidents, hazardous chemical leaks, natural disasters, and social rescue. Faced with a large amount of emergency information, fire command centers need to quickly and accurately analyze and summarize the information to form a summary report, thereby providing support for commanders to make informed decisions.
[0003] Existing technologies mostly rely on manual summarization of massive amounts of police information to generate police report summaries. However, this method is prone to omissions and biases, severely undermining the accuracy of police report assessments. Summary of the Invention
[0004] To address the aforementioned technical problems, this application provides a fire alarm handling method and device, a computer program product, and a storage medium, which can avoid omissions and deviations, and improve the accuracy of alarm judgment and command and dispatch.
[0005] Firstly, this application provides a method for handling fire alarms, comprising: receiving original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information, and third original alarm information; analyzing the original alarm information according to a fire knowledge base, and integrating the analyzed original alarm information to obtain alarm information to be processed; obtaining alarm tags according to a large language model, a preset keyword system in the fire knowledge base, and the alarm information to be processed; performing key tag identification and risk assessment on the alarm tags to obtain an alarm information list; determining prompt words according to the alarm information list, the frequency of occurrence of key tags, and mutual exclusivity; and generating an alarm summary report according to the large language model and the prompt words.
[0006] The above fire alarm handling method receives multiple alarm information, including first, second, and third original alarm information, and analyzes and integrates them using a fire safety knowledge base to form alarm information to be processed. Next, combining a large language model, the keyword system in the fire safety knowledge base, and the alarm information to be processed, alarm tags are accurately extracted, and key tag identification and risk assessment are performed to generate an alarm information list. Furthermore, based on the alarm information list, the frequency of occurrence of key tags, and their mutual exclusivity, prompt words are determined. Finally, a summary report of the alarm situation is generated using the large language model and prompt words. This method not only achieves automated processing and accurate summarization of alarm information but also avoids omissions and biases, improving the accuracy of alarm judgment and command and dispatch. It provides strong support for fire command decision-making and effectively improves the efficiency and scientific nature of fire alarm handling.
[0007] In one implementation, the second original alarm information is obtained by real-time transcription of the alarm telephone voice, and the third original alarm information is obtained by real-time transcription of radio voice information. The original alarm information is analyzed according to the fire protection knowledge base, specifically including: extracting information from the original alarm information according to the preset keyword system in the fire protection knowledge base to obtain target keywords; when the original alarm information is the second original alarm information and / or the third original alarm information, semantic correction and contextual reasoning completion are performed on the second original alarm information and / or the third original alarm information.
[0008] One implementation also includes removing unusable information when it exists in the original alarm information.
[0009] The above fire alarm handling methods, when analyzing the original alarm information, extract information based on the keyword system in the fire safety knowledge base to accurately obtain target keywords. For the second and third original alarm information, semantic correction and contextual reasoning completion steps are added to ensure the accuracy and completeness of the alarm information. Furthermore, when unusable information is found in the original alarm information, it is removed, further improving the quality of alarm information acquisition. This also further improves the accuracy of generating alarm summary reports and helps fire commanders quickly grasp the full picture of the alarm, enhancing the scientific nature of emergency decision-making.
[0010] In one implementation, key labels include at least one of fire alarm level, geographical location, importance of the burning object, and response force.
[0011] One implementation also includes: synchronously updating the temporary tag library in the fire safety knowledge base based on the alarm tag; and synchronously updating the prompt word library and template library based on the prompt words.
[0012] In one implementation, the method further includes: preprocessing the original alarm information and performing similarity matching based on the vector database; and integrating the original alarm information processed by the vector database and the fire knowledge base to obtain the alarm information to be processed.
[0013] Secondly, this application also provides a fire alarm processing device, comprising: a receiving module configured to receive original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information, and third original alarm information; an analysis module configured to: analyze the original alarm information according to a fire knowledge base, and integrate the analyzed original alarm information to obtain alarm information to be processed; obtain alarm tags according to a large language model, a preset keyword system in the fire knowledge base, and the alarm information to be processed; perform key tag identification and risk assessment on the alarm tags to obtain an alarm information list; determine prompt words according to the alarm information list, the frequency of occurrence of key tags, and mutual exclusivity; and a generation module configured to generate an alarm summary report according to the large language model and the prompt words.
[0014] In one implementation, the second original alarm information is obtained by real-time voice transcription of the alarm telephone, and the third original alarm information is obtained by real-time voice transcription of radio information; the analysis module is also configured to: extract information from the original alarm information according to the preset keyword system in the fire knowledge base to obtain target keywords; when the original alarm information is the second original alarm information and / or the third original alarm information, perform semantic correction and contextual reasoning completion on the second original alarm information and / or the third original alarm information.
[0015] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described fire alarm handling methods.
[0016] Fourthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described fire alarm handling methods.
[0017] Compared with the prior art, the present invention has at least one of the following beneficial effects: 1. By receiving various alarm information, including first, second, and third original alarm information, the system analyzes and integrates this information using a fire safety knowledge base to form a list of alarms to be processed. Next, combining a large language model, the keyword system in the fire safety knowledge base, and the alarm information to be processed, alarm tags are accurately extracted, and key tag identification and risk assessment are performed to generate an alarm information list. Furthermore, based on the alarm information list, the frequency of occurrence of key tags, and their mutual exclusivity, prompt words are determined. Finally, a summary report of the alarms is generated using the large language model and prompt words. This method not only automates the processing and accurate summarization of alarm information but also avoids omissions and biases, improving the accuracy of alarm judgment and command and dispatch. It provides strong support for fire command decision-making and effectively improves the efficiency and scientific nature of fire alarm handling.
[0018] 2. When analyzing the original emergency call information, information is extracted based on the keyword system in the fire safety knowledge base to accurately obtain target keywords. For the second and third original emergency call information, semantic correction and contextual reasoning completion steps are added to ensure the accuracy and completeness of the information. Furthermore, when unusable information is found in the original emergency call information, it is removed, further improving the quality of information acquisition. This also improves the accuracy of the generated emergency call summary report and helps fire commanders quickly grasp the full picture of the emergency, enhancing the scientific nature of emergency decision-making. Attached Figure Description
[0019] The preferred embodiments will now be described in a clear and easy-to-understand manner, in conjunction with the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of the present invention.
[0020] Figure 1 This paper shows a schematic diagram of the structure of a fire alarm handling system provided in an embodiment of this application; Figure 2 A flowchart of a fire alarm handling method provided in an embodiment of this application is shown; Figure 3 This document illustrates a flowchart of an embodiment of the present application for analyzing original police information. Figure 4 A structural block diagram of a fire alarm handling device provided in an embodiment of this application is shown. Detailed Implementation
[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.
[0022] To keep the drawings concise, each figure only schematically shows the parts relevant to the invention, and these do not represent the actual structure of the product. Furthermore, to facilitate understanding, in some figures, only one of components with the same structure or function is schematically depicted, or only one is labeled. In this document, "one" not only means "only one," but can also mean "more than one."
[0023] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0024] In this document, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0025] Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0026] It should be noted that the above embodiments can be freely combined as needed. The above are merely preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
[0027] Emergency alert information refers to critical information closely related to rescue operations, collected, organized, and transmitted through various channels during fires or other disasters. It encompasses numerous elements, including the time, location, and type of the incident (e.g., fire, earthquake, chemical leak), the affected area, casualties, on-site hazards (e.g., flammable and explosive materials, toxic gases), and preliminary rescue measures already taken. This information plays a vital role in enabling fire departments to quickly and accurately formulate rescue strategies, efficiently allocate rescue resources, and smoothly implement rescue operations. Therefore, in-depth analysis and assessment of emergency alert information, resulting in a summary report, is crucial. This summary report helps relevant commanders make rapid and accurate scientific judgments about specific incidents, providing a strong basis for subsequent command and dispatch work, thereby ensuring the efficient and orderly conduct of rescue operations.
[0028] Some methods for summarizing police incidents utilize databases for keyword searches to achieve this. However, when faced with diverse, heterogeneous, and large-volume police incident records, this method struggles to generate accurate police incident summary reports.
[0029] Other methods of summarizing and analyzing police reports rely on manual summarization of massive amounts of information within a short period. However, this approach is highly susceptible to omissions and biases, which not only weakens the accuracy of incident assessment but may also cause missed opportunities for optimal command and dispatch. Especially in sudden, large-scale fires or disasters, information overload is a key factor leading to a significant decrease in analytical efficiency.
[0030] Large-scale pre-trained language models (LMMs) are complex neural network models built on deep learning technology. Through unsupervised learning on massive amounts of text data, they can automatically learn the grammatical structure, semantic information, and contextual relationships of a language. In the field of fire alarm information processing, LLMs can efficiently summarize complex text, accurately extract key elements, and automatically generate fire alarm summary reports. Compared with traditional methods, LLMs can not only automatically process unstructured text but also uncover implicit patterns in multi-turn semantic reasoning, thereby significantly improving the efficiency and accuracy of alarm summary.
[0031] Therefore, this application provides a method for handling fire alarms, which uses a large language model to analyze alarm information and obtain an alarm summary report. This can avoid omissions and biases, and improve the accuracy of alarm judgment and command and dispatch.
[0032] The following explanation is based on the accompanying diagram: Reference Appendix Figure 1 This illustration shows a structural diagram of a fire alarm handling system provided in an embodiment of this application. Figure 1As shown, it includes: an alarm handling system, responsible for receiving alarms, dispatching, and inputting subsequent feedback information, serving as one of the main service data sources; a two-way radio system, used to provide radio dialogue voice streams for alarm situations, as well as most of the instruction records and on-site situation voice descriptions after alarm dispatch, also serving as one of the main service data sources; a speech-to-text module, used to realize the speech-to-text function of alarm recordings, callback inquiry calls, and radio recordings; a fire safety knowledge base, used to provide commonly used fire safety tags, force information, personnel information, and other knowledge data, providing a baseline for alarm information processing; and a large language model, used to extract information from pre-processed alarm information, call records, radio records, etc. The system extracts keywords and generates a summary report of the incident based on prompts and a summary template. The data preprocessing module utilizes a vector database to perform preliminary data preprocessing of incident information and similarity matching of existing tags. It performs real-time format conversion, deduplication, and removal of invalid information on information generated during firefighting and rescue operations to ensure data standardization and consistency. The prompt engineering module compares the vector database to identify key tags from keywords such as fire alarm level, geographical location, importance of the burning object, and dispatched forces. Combining the information list with the frequency and mutual exclusivity of key tags, it generates targeted prompts for each incident summary.
[0033] This embodiment of the application integrates an alarm handling system, an intercom system, a speech-to-text module, a fire safety knowledge base, a large language model, a data preprocessing module, and a prompt word engineering module to achieve full automation and intelligence throughout the entire process from receiving, processing, to summarizing alarm information. The alarm handling system and intercom system serve as the main data sources, ensuring comprehensive collection of alarm information; the speech-to-text module converts speech information into text, improving information processing efficiency; the fire safety knowledge base provides rich knowledge support for alarm handling; the large language model can accurately extract keywords and generate alarm summary reports based on prompt words, improving the accuracy and efficiency of alarm summary; the data preprocessing module uses a vector database to standardize and ensure data consistency, ensuring data quality; and the prompt word engineering module generates targeted prompt words based on key tags, further improving the accuracy and relevance of alarm summaries. The collaborative work of these modules not only improves the efficiency and accuracy of fire alarm summary but also provides strong support for fire command and decision-making, effectively enhancing the overall effectiveness of fire rescue.
[0034] Reference Appendix Figure 2 The document illustrates a flowchart of a fire alarm handling method provided in an embodiment of this application. This method can be applied to the aforementioned fire alarm handling system, such as... Figure 2 As shown, it includes: S200, receive original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information and third original alarm information.
[0035] S210: Analyze the original alarm information based on the fire safety knowledge base, and integrate the analyzed original alarm information to obtain the alarm information to be processed.
[0036] S220 generates alarm tags based on the large language model, the preset keyword system in the fire protection knowledge base, and the alarm information to be processed.
[0037] S230 performs key label identification and risk assessment on the police incident labels to obtain a list of police incident information.
[0038] S240: Determine prompt words based on the list of police information, the frequency of occurrence of key tags, and their mutual exclusivity.
[0039] S250 generates a summary report of police incidents based on a large language model and prompt words.
[0040] The first type of original alarm information can be the case information (or alarm text information) received by the alarm handling system. The second type of original alarm information can be the alarm telephone information received by the alarm handling system, which is then transcribed in real time by the voice transcription module. The third type of original alarm information can be the radio voice information received by the walkie-talkie system, which is then transcribed in real time by the voice transcription module. The specific voice transcription process is as follows: after receiving the alarm telephone information and / or radio voice information, the voice transcription module uses ASR (Automatic Speech Recognition) technology to transcribe the content of the alarm telephone information and / or radio voice information in real time. This process of transcribing while the call is being made can capture alarm information in a timely manner. The transcribed text will serve as the basis for further analysis, helping the dispatcher to understand the situation on the scene more quickly and improving response speed.
[0041] Upon receiving the original emergency alert information, it can be analyzed using keyword recognition rules from the fire safety knowledge base. This analysis may include, but is not limited to, keyword extraction, semantic correction, and contextual reasoning completion. For example, if the original alert information is at least one of the first, second, and third types, information extraction can be performed, focusing on extracting keywords related to core elements such as time, location, people, nature of the event, and urgency. Similarly, if the original alert information is the second and / or third type, semantic correction and contextual reasoning completion can be performed. Then, the analyzed original alert information can be integrated into a complete alert information set for processing.
[0042] The information of the emergency calls to be processed is input into a large language model. Based on a pre-defined keyword system in the fire safety knowledge base, the large language model can automatically extract and pre-process information from the unstructured text of the emergency calls, focusing on the following: emergency type (e.g., fire, rear-end collision, drowning), attributes of the involved object (e.g., wasp nest, door lock, bicycle), object location (e.g., tunnel, shop, non-motorized vehicle lane), special emergency status (e.g., injuries, trapped persons, gas odor, etc.), contact information of the caller (extracting valid phone numbers or communication identifiers), smoke conditions (e.g., charring, burnt, thick smoke), trapped persons (e.g., unlocking, opening, breaking down doors), and key attention tags (e.g., open flame, sparks, fire). This allows the large language model to output emergency tags.
[0043] During the process of inputting alarm information into the big data language model and outputting alarm labels, additional processing steps can be performed, such as deduplication and filtering of invalid content (e.g., interjections, repetitions, and irrelevant dialogue). Combined with semantic similarity analysis, it can be determined whether the current alarm corresponding to the alarm information to be processed is a duplicate alarm or a false alarm, improving the accuracy of alarm identification and the efficiency of handling. The judgment result is then communicated to the user through the alarm handling system.
[0044] After obtaining the alarm tags, the prompt word engineering module can perform key tag identification (such as identifying keywords in the key tags such as fire alarm level, geographical location, importance of the burning object, and dispatch force) and risk assessment. For example, through preset semantic rules and risk level models, keywords with clear physical hazard characteristics in the key tags can be monitored and identified, such as "ignited," "gas cylinder," "black smoke," "charred," "fire," and "open flame," all of which are classified as high-risk signals of fire or potential fire risk.
[0045] At the same time, we will also be highly vigilant about keywords (or sensitive tags) in key labels that involve personnel safety and psychological crisis. For example, expressions such as "salary issues", "emotional agitation", "sleeping pills", "sitting on the roof", and "jumping off a building" will be intelligently identified as risk signals that are highly associated with "suicide", "suicidal tendencies" or "extreme behavior".
[0046] After completing key tag identification and risk assessment, a structured and standardized list of police incident information can be automatically generated. Through this intelligent process, rapid extraction of police incident information and early warning of risks can be achieved, enabling efficient allocation of resources and comprehensively improving the response speed and handling efficiency of public safety incidents.
[0047] The prompt word engineering module combines the frequency and mutual exclusivity of the police incident information list and key tags to generate targeted prompt words for each police incident summary. The large language model can then generate a police incident summary report based on the generated prompt words and summary template. The generation of this summary report is real-time; that is, during the handling of a police incident, the emergency response system and radio system may generate new case information, emergency call information, and radio voice information at any time. Once this information is collected and converted into original police incident information, the aforementioned steps must be executed to generate a new police incident summary report.
[0048] The police incident summary report includes: automatically generated prompts and output labels, manually verified final labels, and the prompts used by dispatchers or commanders during actual operations. By calculating the difference between the model output and actual usage, comparative analysis is conducted to optimize the prompting engineering strategy, thereby continuously improving the semantic understanding accuracy and practical application effectiveness of the model in real policing scenarios.
[0049] This application embodiment receives multiple alarm information, including first, second, and third original alarm information, and analyzes and integrates them using a fire safety knowledge base to form alarm information to be processed. Then, combining a large language model, the keyword system in the fire safety knowledge base, and the alarm information to be processed, alarm tags are accurately extracted, and key tag identification and risk assessment are performed to generate an alarm information list. Further, based on the alarm information list, the frequency of occurrence of key tags, and their mutual exclusivity, prompt words are determined. Finally, a summary report of the alarm situation is generated using the large language model and the prompt words. This method not only achieves automated processing and accurate summarization of alarm information but also avoids omissions and deviations, improving the accuracy of alarm judgment and command and dispatch. It provides strong support for fire command decision-making and effectively improves the efficiency and scientific nature of fire alarm handling.
[0050] After obtaining the original alarm information, it can be preprocessed to remove invalid information, ensure data consistency and standardization, and further improve the accuracy of alarm summary reports. For example, in some embodiments of this application, it further includes: preprocessing the original alarm information and performing similarity matching based on a vector database; and integrating the original alarm information processed by the vector database and fire knowledge base to obtain alarm information to be processed.
[0051] After obtaining the original alarm information, it can be analyzed using a fire safety knowledge base. Simultaneously, preprocessing and similarity matching can be performed on the original alarm information using a vector database. For example, a vector database can be used to perform real-time format conversion, deduplication, and removal of invalid information from information generated during firefighting and rescue operations.
[0052] Reference Appendix Figure 3 This illustrates a flowchart of an embodiment of the present application for analyzing original police information. Figure 3 As shown, it includes: S300 extracts target keywords from the original alarm information based on the preset keyword system in the fire knowledge base.
[0053] S310, when the original alarm information is the second original alarm information and / or the third original alarm information, perform semantic correction and contextual reasoning completion on the second original alarm information and / or the third original alarm information.
[0054] After obtaining the original police report information, keyword extraction, semantic correction, contextual reasoning to complete the information, and elimination of unusable information can be performed. For example, when the original police report information is at least one of the first, second, and third types, information extraction can be performed to extract keywords related to core elements such as time, location, people, nature of the event, and urgency. For example, the identity of the person reporting the incident can be confirmed by using words like "passing by" or "property management"; the frequency of "inside the house" versus "outside the house" can be compared to determine whether the incident occurred indoors or outdoors; the nature of the event can be determined by words like "open flame" and "drowning"; and "open flame" and "smoke" are associated with fire alarms and require special attention.
[0055] For example, when the original police report is the second and / or third original police report, semantic correction and contextual reasoning can be performed on the original police report. For example, "futures" is actually "fire", "Mo Yan" is actually "smoke", "poisonous tongue" and "sticking out tongue" are actually "poisonous snake", "owing money" and "breaking up" are associated with "jumping off a building" and "suicide", "extinguishing" is marked as handled, "pumping water" is associated with drainage in conjunction with keywords such as water accumulation, heavy rain, and flooding, and in the absence of fire and smoke, "cat" and "dog" are associated with social assistance, "being crushed" is associated with people trapped in equipment, and "cannot get in" is associated with emergency door opening.
[0056] For example, when the original alarm information is at least one of the first, second, and third types of original alarm information, the usability of the original alarm information can be analyzed in conjunction with the fire safety knowledge base. For instance, in an alarm handling scenario, if a casual chat conversation such as "What did you eat?" is mixed in and has no relation to the alarm handling, it will be judged as useless information. However, if it is related to the alarm handling, such as "People are trapped", it will be integrated together.
[0057] Finally, the raw alarm information obtained from the aforementioned analysis process can be integrated into a complete alarm information to be processed. Then, steps S220 to S250 can be executed on the alarm information to be processed.
[0058] In analyzing the original alarm information, this embodiment extracts information based on the keyword system in the fire safety knowledge base to accurately obtain target keywords. For the second and third original alarm information, semantic correction and contextual reasoning completion steps are added to ensure the accuracy and completeness of the alarm information. Furthermore, when unusable information exists in the original alarm information, it is removed, further improving the quality of alarm information acquisition. This also further improves the accuracy of generating alarm summary reports and helps fire commanders quickly grasp the full picture of the alarm situation, enhancing the scientific nature of emergency decision-making.
[0059] In one embodiment of this application, reference continues to the appendix. Figure 2 It also includes: S260, which updates the temporary tag library in the fire safety knowledge base in sync with the alarm tags, and updates the prompt word library and template library in sync with the prompt words.
[0060] Each time a summary report of an incident is generated based on the integrated information of the incidents to be processed, the system continuously optimizes the tagging system and reasoning logic, and dynamically iterates the "temporary key prompt word library" and the "handling report generation template library." For example, the incident tags output by the large model can be used to synchronously update the temporary tag library in the knowledge base. The tagging system and its associated logic in the knowledge base will be regularly evaluated and analyzed, and the tag definitions and reasoning rules will be dynamically adjusted and optimized accordingly to achieve continuous evolution and iterative upgrades of the knowledge system. The system will also dynamically update the "temporary key prompt word library" and the "handling report generation template library" based on prompt words (or based on high-frequency incident patterns and the optimized knowledge structure). Updating the temporary key prompt word library helps improve the sensitivity of the large language model to emerging incident characteristics, while iterating the template library can generate more targeted, standardized, and operational handling suggestions, thereby comprehensively improving the intelligence level and practical effectiveness of incident analysis.
[0061] Reference Appendix Figure 4 The diagram illustrates a structural block diagram of a fire alarm handling device provided in an embodiment of this application. Figure 4As shown, the device 400 includes: a receiving module 410 configured to receive original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information, and third original alarm information; an analysis module 420 configured to: analyze the original alarm information according to a fire protection knowledge base, and integrate the analyzed original alarm information to obtain alarm information to be processed; obtain alarm tags according to a large language model, a preset keyword system in the fire protection knowledge base, and the alarm information to be processed; perform key tag identification and risk assessment on the alarm tags to obtain an alarm information list; determine prompt words according to the alarm information list, the frequency of occurrence of key tags, and mutual exclusivity; and a generation module 430 configured to generate an alarm summary report according to the large language model and prompt words.
[0062] In some embodiments of this application, the second original alarm information is obtained by real-time voice transcription of the alarm telephone, and the third original alarm information is obtained by real-time transcription of radio voice information; the analysis module is also configured to: extract information from the original alarm information according to the preset keyword system in the fire knowledge base to obtain target keywords; when the original alarm information is the second original alarm information and / or the third original alarm information, perform semantic correction and contextual reasoning completion on the second original alarm information and / or the third original alarm information.
[0063] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the fire alarm handling method of any of the above embodiments.
[0064] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the fire alarm handling method of any of the above embodiments.
[0065] It should be noted that the above embodiments can be freely combined as needed. The above are merely preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for handling fire alarms, characterized in that, include: Receive original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information and third original alarm information; The original alarm information is analyzed based on the fire safety knowledge base, and the analyzed original alarm information is integrated to obtain the alarm information to be processed. Based on the large language model, the preset keyword system in the fire protection knowledge base, and the alarm information to be processed, alarm tags are obtained; The police incident tags are subjected to key tag identification and risk assessment to obtain a list of police incident information; Based on the list of police information, the frequency of occurrence of key tags, and their mutual exclusivity, the prompt words are determined; Based on the large language model and the prompt words, a police incident summary report is generated.
2. The fire alarm handling method according to claim 1, characterized in that, The second original alarm information was obtained by real-time transcription of the alarm telephone voice, and the third original alarm information was obtained by real-time transcription of the radio voice information; The analysis of the original alarm information based on the fire safety knowledge base specifically includes: Based on the preset keyword system in the fire safety knowledge base, information is extracted from the original alarm information to obtain target keywords; When the original alarm information is the second original alarm information and / or the third original alarm information, semantic correction and contextual reasoning completion are performed on the second original alarm information and / or the third original alarm information.
3. The fire alarm handling method according to claim 2, characterized in that, Also includes: If there is unusable information in the original police information, the unusable information will be removed.
4. The fire alarm handling method according to claim 1, characterized in that, The key labels include at least one of the following: fire alarm level, geographical location, importance of the burning object, and response force.
5. The fire alarm handling method according to claim 1, characterized in that, Also includes: Based on the aforementioned alarm tags, the temporary tag library in the fire safety knowledge base is updated synchronously. Based on the provided prompt words, the prompt word library and template library are updated synchronously.
6. The fire alarm handling method according to any one of claims 1-5, characterized in that, Also includes: The original police information is preprocessed and similarity matched based on the vector database; The original alarm information, after being processed by the vector database and the fire knowledge base, is integrated to obtain the alarm information to be processed.
7. A fire alarm handling device, characterized in that, include: The receiving module is configured to receive original alarm information, wherein the original alarm information includes at least one of first original alarm information, second original alarm information, and third original alarm information; The analysis module is configured to: analyze the original alarm information based on the fire safety knowledge base, and integrate the analyzed original alarm information to obtain alarm information to be processed; Based on the large language model, the preset keyword system in the fire protection knowledge base, and the alarm information to be processed, alarm tags are obtained; key tag identification and risk assessment are performed on the alarm tags to obtain an alarm information list; prompt words are determined based on the alarm information list, the frequency of occurrence of key tags, and mutual exclusivity. The generation module is configured to generate a police situation summary report based on the large language model and the prompt words.
8. The fire alarm handling device according to claim 7, characterized in that, The second original alarm information was obtained by real-time transcription of the alarm telephone voice, and the third original alarm information was obtained by real-time transcription of the radio voice information; The analysis module is further configured to: extract information from the original alarm information according to the preset keyword system in the fire protection knowledge base to obtain target keywords; and when the original alarm information is the second original alarm information and / or the third original alarm information, perform semantic correction and contextual reasoning completion on the second original alarm information and / or the third original alarm information.
9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the fire alarm handling method according to any one of claims 1-6.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the fire alarm handling method according to any one of claims 1-6.