Generative intelligent dialogue methods, systems, equipment, and media for rainstorm and flood disaster avoidance.

Abstract

This invention relates to the field of intelligent emergency management technology, and particularly to a generative intelligent dialogue method, system, device, and medium for rainstorm and flood avoidance. The method includes: loading an expert network model under a dynamic hybrid expert architecture based on the operating status of a portable edge device; preprocessing voice stream data; predicting and completing interrupted voice content in the voice stream data; uploading crowdsourced data of the user's location; assimilating the verified crowdsourced data into a preset flood forecast model; outputting the confidence score of standard text; generating user semantic information and environmental data that meet preset high confidence conditions; and generating a dynamic rainstorm and flood avoidance dialogue strategy based on the flood and rainstorm warning level and the user's location coordinates. This invention achieves adaptive interaction and dynamic allocation of edge device computing resources in flood scenarios, and uses crowdsourced data from community interaction to assimilate and dynamically correct the flood model in real time, ensuring the safety of life and property for the elderly in the face of extreme flood weather.

Description

Technical Field

[0001] This invention relates to the field of intelligent emergency management technology, and in particular to a generative intelligent dialogue method, system, device and medium for rainstorm and flood disaster avoidance. Background Technology

[0002] The development of voice interaction technology has provided new ideas for improving disaster response capabilities for the elderly. In recent years, smart devices (such as smartphones and smart speakers) have been able to provide assistance and guidance through voice dialogue. However, there are still technical challenges in practical applications: First, insufficient dialect recognition capabilities. The elderly rely heavily on dialects, and if the system cannot recognize their local dialect, they will not be able to understand distress messages; second, the discontinuous nature of speech makes it difficult for the elderly to communicate smoothly with peripheral devices; third, the high environmental noise and communication interruption risk in flood disaster scenarios require voice systems to have high robustness and adaptability.

[0003] Existing large-scale models such as ChatGPT (Chat Generative Pre-trained Transformer) support multi-turn dialogues, but their high-frequency calls are extremely costly. During severe floods, they may trigger millions of interaction requests per day, easily causing a "bloat" of computing resources. The Claude model has an excessively long training period for dialect data, making it difficult to meet the emergency management requirement of "system readiness within 30 minutes before a disaster." Therefore, handling high-concurrency interactions on edge devices with limited computing power is a pressing technical challenge that needs to be addressed.

[0004] Furthermore, traditional flood forecasting relies primarily on fixed monitoring stations, resulting in limited spatial coverage. Crowdsourced assimilation experiments have demonstrated that public on-site observation can significantly compensate for information gaps caused by sparse instrumentation, especially in urban flooding and rural flash flood scenarios. However, current intelligent interactive systems have not yet integrated the "community interaction-data assimilation-dynamic correction" closed loop to the edge, causing a disconnect between forecasts and actual flood conditions, which urgently needs improvement. Summary of the Invention

[0005] This invention provides a generative intelligent dialogue method, system, device, and medium for rainstorm and flood disaster avoidance. It addresses the problems of existing large-scale models having extremely high costs due to frequent calls, potentially triggering millions of interaction requests daily during rainstorm disasters, easily causing a "burst" of computing resources; the Claude model having an excessively long training period for dialect data, making it difficult to meet the emergency management requirement of "system readiness within 30 minutes before disaster"; traditional flood forecasting mainly relying on fixed monitoring stations with limited spatial coverage; and current intelligent interactive systems not yet integrating the "community interaction-data assimilation-dynamic correction" closed loop to the edge, resulting in a disconnect between forecasts and actual flood conditions.

[0006] A first aspect of this invention provides a generative intelligent dialogue method for flood and rainstorm disaster avoidance, comprising the following steps: initializing a prefix tree data structure and loading an expert network model under a dynamic hybrid expert architecture based on the operating status of a portable edge device; utilizing the expert network model to acquire the user's voice stream data in real time based on the portable edge device when the user is in a flood disaster emergency interaction scenario, and preprocessing the voice stream data to extract voice features that meet preset clarity conditions; analyzing the voice features that meet the preset clarity conditions to predict and complete the interrupted voice content of the voice stream data, generating a complete voice sequence; uploading crowdsource data of the user's location and verifying the crowdsource data. The system uses data to assimilate validated crowdsourced data into a preset flood forecasting model, generating corrected flood level and flow forecasts, and generating flood risk information based on these forecasts. Based on this flood risk information, it decodes dialect speech in the complete speech sequence into standard text and outputs the confidence level of the standard text. Based on the confidence level of the standard text, it asks questions to the user to guide them in clarifying ambiguous semantics, generating user semantic information and environmental data that meet preset high confidence conditions. Based on the user semantic information and environmental data that meet the preset high confidence conditions, it generates a dynamic flood and rainstorm avoidance dialogue strategy according to the flood and rainstorm warning level and the user's location coordinates.

[0007] Optionally, in one embodiment of the present invention, the step of loading the expert network model under the dynamic hybrid expert architecture according to the operating status of the portable edge device includes: determining the dialect recognition expert model, semantic clarification expert model, risk avoidance decision expert model, equipment optimization expert model, and data assimilation and alignment expert model in the flood disaster expert network model according to the operating status of the portable edge device; simultaneously loading the dialect recognition expert model, the semantic clarification expert model, the risk avoidance decision expert model, the equipment optimization expert model, and the data assimilation and alignment expert model when the remaining power of the portable edge device is greater than a first preset threshold; loading the equipment optimization expert model when the remaining power is greater than a second preset threshold and less than or equal to the first preset threshold; and loading the dialect recognition expert model, the semantic clarification expert model, and the risk avoidance decision expert model when the remaining power is less than or equal to the second preset threshold.

[0008] Optionally, in one embodiment of the present invention, after acquiring the user's voice stream data in real time based on the portable edge device, the method further includes: if a preset keyword is detected in the voice stream data, pausing non-critical background tasks and increasing the weight of resource allocation to a preset percentage; if a modifier in the voice stream data is detected to be greater than a preset threshold, calling a prefix tree algorithm to analyze the syntax in order to generate the core semantic components of the statement.

[0009] Optionally, in one embodiment of the present invention, the step of analyzing the speech features that meet the preset clarity conditions to predict and complete the interrupted speech content of the speech stream data and generate a complete speech sequence includes: when a long-term interruption or blurred phoneme segment is detected in the speech stream data, using the prefix tree algorithm to trigger a node backtracking mechanism to count the phoneme path frequency of the speech features; backtracking to select a phoneme sequence path based on the phoneme path frequency, so as to predict and complete the interrupted speech content based on the phoneme sequence path and generate the complete speech sequence.

[0010] Optionally, in one embodiment of the present invention, the step of uploading crowdsourced data of the user's location and verifying the crowdsourced data to assimilate the verified crowdsourced data into a preset flood forecasting model to generate corrected flood level and flow forecasts, and generating flood risk information based on the corrected flood level and flow forecasts, includes: uploading crowdsourced data of the user's location and preprocessing the crowdsourced data to generate preprocessed crowdsourced data; cross-validating the preprocessed crowdsourced data to generate verified crowdsourced data; assimilating the verified crowdsourced data into the preset flood forecasting model to generate the corrected flood level and flow forecast data; and based on the corrected flood level and flow forecast data, identifying the magnitude of flood risk changes, and adjusting the flood threshold and warning level according to the magnitude of flood risk changes to generate the flood risk information.

[0011] Optionally, in one embodiment of the present invention, the credibility calculation formula for the standard text is:

[0012]

[0013] in, C This represents the confidence value of the spatiotemporal alignment of multi-source data in the system. Indicates the weight of the time dimension. Indicates the spatial dimension weight. Represents the semantic dimension weight. This indicates the credibility score corresponding to the time dimension. This represents the credibility score corresponding to the spatial dimension. This represents the credibility score corresponding to the semantic dimension. Indicates a time-dimensional index. Indicates spatial dimension index, Represents a semantic dimension index;

[0014] The formula for calculating the credibility score corresponding to the time dimension is as follows:

[0015]

[0016] in, This represents the absolute difference between the time of voice interaction and the time recorded by the sensor. This indicates the maximum permissible time difference under the warning level;

[0017] The formula for calculating the credibility score corresponding to the spatial dimension is as follows:

[0018]

[0019] in, Indicates the user's location and the first i The distance between monitoring points Indicates the first i Water level status at each monitoring point Indicates the monitoring point index. Indicates the number of monitoring points;

[0020] The formula for calculating the credibility score corresponding to the semantic dimension is as follows:

[0021]

[0022] in, n This indicates the number of times a question was asked to clarify the semantic meaning.

[0023] A second aspect of this invention provides a generative intelligent dialogue system for rainstorm and flood disaster avoidance, comprising: a prefix tree initialization module, used to initialize the data structure of the prefix tree and load an expert network model under a dynamic hybrid expert architecture according to the operating status of a portable edge device; a preprocessing module, used to utilize the expert network model to acquire the user's voice stream data in real time according to the portable edge device when the user is in a flood disaster emergency interaction scenario, and preprocess the voice stream data to extract voice features that meet preset clarity conditions; a prediction module, used to analyze the voice features that meet the preset clarity conditions to predict and complete the interrupted voice content of the voice stream data, generating a complete voice sequence; and a crowdsourced data upload module, used to upload crowdsourced data of the user's location and verify... The system verifies the crowdsourced data, assimilates the verified crowdsourced data into a preset flood forecast model, generates corrected flood level and flow forecasts, and generates flood risk information based on the corrected flood level and flow forecasts; a decoding module decodes the dialect speech in the complete speech sequence into standard text based on the flood risk information, and outputs the confidence level of the standard text; an interaction module asks questions to the user based on the confidence level of the standard text to guide the user to clarify ambiguous semantics, and generates user semantic information and environmental data that meet preset high confidence conditions; a risk avoidance dialogue module generates a dynamic risk avoidance dialogue strategy for rainstorms and floods based on the user semantic information that meets preset high confidence conditions and the environmental data, according to the flood and rainstorm warning level and the user's location coordinates.

[0024] Optionally, in one embodiment of the present invention, the prefix tree initialization module includes: a determination unit, configured to determine, based on the operating status of the portable edge device, a dialect recognition expert model, a semantic clarification expert model, a risk avoidance decision expert model, a device optimization expert model, and a data assimilation and alignment expert model in the flood disaster expert network model; a first loading unit, configured to simultaneously load the dialect recognition expert model, the semantic clarification expert model, the risk avoidance decision expert model, the device optimization expert model, and the data assimilation and alignment expert model when the remaining power of the portable edge device is greater than a first preset threshold; a second loading unit, configured to load the device optimization expert model when the remaining power is greater than a second preset threshold and less than or equal to the first preset threshold; and a third loading unit, configured to load the dialect recognition expert model, the semantic clarification expert model, and the risk avoidance decision expert model when the remaining power is less than or equal to the second preset threshold.

[0025] Optionally, in one embodiment of the present invention, it further includes: a pause module, configured to pause non-critical background tasks and increase the weight of resource allocation to a preset percentage when a preset keyword is detected in the voice stream data after the user's voice stream data is acquired in real time by the portable edge device; and a call module, configured to call a prefix tree algorithm to analyze the syntax to generate the core semantic components of the statement when a modifier in the voice stream data is detected to be greater than a preset threshold.

[0026] Optionally, in one embodiment of the present invention, the prediction module includes: a statistics unit, used to trigger a node backtracking mechanism using the prefix tree algorithm to count the phoneme path frequency of the speech features when a long-term interruption or ambiguous phoneme segment is detected in the speech stream data; and a prediction unit, used to backtrack and select a phoneme sequence path based on the phoneme path frequency to predict and complete the interrupted speech content based on the phoneme sequence path, thereby generating the complete speech sequence.

[0027] Optionally, in one embodiment of the present invention, the crowdsourced data upload module includes: an upload unit for uploading crowdsourced data of the user's location and preprocessing the crowdsourced data to generate preprocessed crowdsourced data; a cross-validation unit for cross-validating the preprocessed crowdsourced data to generate validated crowdsourced data; an assimilation unit for assimilating the validated crowdsourced data to the preset flood forecast model to generate the corrected flood level and flow forecast data; and an adjustment unit for identifying the magnitude of flood risk change based on the corrected flood level and flow forecast data, and adjusting the flood threshold and warning level according to the magnitude of flood risk change to generate the flood risk information.

[0028] Optionally, in one embodiment of the present invention, the credibility calculation formula for the standard text is:

[0029]

[0030] in, C This represents the confidence value of the spatiotemporal alignment of multi-source data in the system. Indicates the weight of the time dimension. Indicates the spatial dimension weight. Represents the semantic dimension weight. This indicates the credibility score corresponding to the time dimension. This represents the credibility score corresponding to the spatial dimension. This represents the credibility score corresponding to the semantic dimension. Indicates a time-dimensional index. Indicates spatial dimension index, Represents a semantic dimension index;

[0031] The formula for calculating the credibility score corresponding to the time dimension is as follows:

[0032]

[0033] in, This represents the absolute difference between the time of voice interaction and the time recorded by the sensor. This indicates the maximum permissible time difference under the warning level;

[0034] The formula for calculating the credibility score corresponding to the spatial dimension is as follows:

[0035]

[0036] in, Indicates the user's location and the first i The distance between monitoring points Indicates the first i Water level status at each monitoring point Indicates the monitoring point index. Indicates the number of monitoring points;

[0037] The formula for calculating the credibility score corresponding to the semantic dimension is as follows:

[0038]

[0039] in, n This indicates the number of times a question was asked to clarify the semantic meaning.

[0040] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the generative rainstorm and flood intelligent disaster avoidance dialogue method as described in the above embodiments.

[0041] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described generative rainstorm and flood intelligent disaster avoidance dialogue method.

[0042] A fifth aspect of the present invention provides a computer program product that stores a computer program that, when executed by a processor, implements the above-described generative rainstorm and flood intelligent disaster avoidance dialogue method.

[0043] In this embodiment of the invention, the MoE architecture works collaboratively through multiple "expert networks." While ensuring model capacity, it activates only experts relevant to the current task, thereby improving computational efficiency. Based on this dynamic MoE architecture, a flood emergency interaction system for the elderly is designed, deploying expert networks for dialect recognition, semantic clarification, hazard avoidance decision-making, device optimization, and cross-modal alignment on edge devices. During interaction, the system dynamically invokes experts as needed: it balances local computing power and battery life through device-optimized experts, switching to cloud-based collaborative computing when necessary to ensure real-time response under peak interaction volumes; and it utilizes cross-modal alignment experts to fuse sensor data to verify the reliability of voice information, providing elderly users with timely and accurate hazard avoidance guidance. This addresses several issues: the high cost of frequent calls to existing large models, which could trigger millions of interaction requests daily during rainstorm disasters, potentially causing a "burst" of computing resources; the excessively long training period for dialect data in the Claude model, making it difficult to meet the emergency management requirement of "system readiness within 30 minutes before disaster"; the reliance on fixed monitoring stations for traditional flood forecasting, which has limited spatial coverage; and the current intelligent interactive system's failure to integrate the "community interaction-data assimilation-dynamic correction" closed loop to the edge, resulting in a disconnect between forecasts and actual flood conditions.

[0044] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 A flowchart of a generative intelligent dialogue method for rainstorm and flood disaster avoidance provided in an embodiment of the present invention;

[0047] Figure 2 This is a flood expert network diagram based on a dynamic MoE architecture according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the overall system structure according to an embodiment of the present invention;

[0049] Figure 4 A flowchart illustrating the data assimilation process according to an embodiment of the present invention;

[0050] Figure 5 This is an emergency interaction flowchart according to an embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of a generative intelligent dialogue system for rainstorm and flood disaster avoidance provided in an embodiment of the present invention;

[0052] Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention.

[0053] Among them, 10-Generative intelligent rainstorm and flood avoidance dialogue system; 100-Prefix tree initialization module, 200-Preprocessing module, 300-Prediction module, 400-Crowdsource data upload module, 500-Decoding module, 600-Interaction module, 700-Avoidance dialogue module; 701-Memory, 702-Processor, 703-Communication interface. Detailed Implementation

[0054] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0055] The following describes, with reference to the accompanying drawings, a generative intelligent dialogue method, apparatus, device, and medium for rainstorm and flood disaster avoidance according to embodiments of the present invention. Addressing the aforementioned background issues, such as the extremely high cost of high-frequency calls to existing large models (potentially triggering millions of interaction requests daily during rainstorm disasters, easily causing a "burst" of computing resources), the excessively long training period for dialect data in the Claude model (failing to meet the emergency management requirement of "system readiness within 30 minutes before disaster"), the limited spatial coverage of traditional flood forecasting relying primarily on fixed monitoring stations, and the current lack of integration of the "community interaction-data assimilation-dynamic correction" closed loop at the edge, resulting in a disconnect between predictions and actual flood conditions, this invention provides a generative intelligent dialogue method for rainstorm and flood disaster avoidance. In this method, the MoE architecture works collaboratively through multiple "expert networks," activating only experts relevant to the current task while ensuring model capacity, thereby improving computational efficiency. Based on the dynamic MoE architecture, a flood emergency interaction system for the elderly is designed, deploying expert networks for dialect recognition, semantic clarification, disaster avoidance decision-making, equipment optimization, and cross-modal alignment on edge devices. The system dynamically invokes experts on demand during interaction: it can balance local computing power and battery life through device-optimized experts, and switch to cloud-based collaborative computing when necessary to ensure real-time response under peak interaction volume. It can also utilize cross-modal alignment of experts to fuse sensor data and verify the reliability of voice information, providing timely and accurate disaster avoidance guidance for elderly users. This solves problems such as the extremely high cost of high-frequency calls to existing large models (potentially triggering millions of interaction requests daily during rainstorm disasters, easily causing a "burst" of computing resources), the excessively long training period of the Claude model for dialect data, making it difficult to meet the emergency management standard of "system readiness within 30 minutes before disaster," traditional flood forecasting mainly relying on fixed monitoring stations with limited spatial coverage, and the current intelligent interactive system not yet integrating the "community interaction-data assimilation-dynamic correction" closed loop to the edge, resulting in a disconnect between forecasts and actual flood conditions.

[0056] Specifically, Figure 1 This is a flowchart illustrating a generative intelligent dialogue method for rainstorm and flood disaster avoidance provided in an embodiment of the present invention.

[0057] like Figure 1 As shown, the generative intelligent dialogue method for rainstorm and flood disaster avoidance includes the following steps:

[0058] In step S101, the data structure of the prefix tree is initialized, and the expert network model under the dynamic hybrid expert architecture is loaded according to the operating status of the portable edge device.

[0059] In actual implementation, embodiments of the present invention can perform system initialization and loading of expert networks. The flood emergency interaction system is started, the prefix tree data structure is initialized, including information such as phoneme fragments and their frequencies stored in tree nodes, and an expert network model under a dynamic MoE (Mixture-of-Experts) architecture is loaded according to the operating status of the portable edge device.

[0060] Optionally, in one embodiment of the present invention, loading an expert network model under a dynamic hybrid expert architecture based on the operating status of the portable edge device includes: determining a dialect recognition expert model, a semantic clarification expert model, a risk avoidance decision expert model, an equipment optimization expert model, and a data assimilation and alignment expert model in the flood disaster expert network model based on the operating status of the portable edge device; simultaneously loading the dialect recognition expert model, the semantic clarification expert model, the risk avoidance decision expert model, the equipment optimization expert model, and the data assimilation and alignment expert model when the remaining power of the portable edge device is greater than a first preset threshold; loading the equipment optimization expert model when the remaining power is greater than a second preset threshold and less than or equal to the first preset threshold; and loading the dialect recognition expert model, the semantic clarification expert model, and the risk avoidance decision expert model when the remaining power is less than or equal to the second preset threshold.

[0061] It is understandable that, such as Figure 2 As shown, the flood disaster emergency interaction system based on the dynamic MoE architecture in this embodiment of the invention constructs five types of flood expert network models by realizing the dynamic activation of MoE experts: dialect recognition expert model, semantic clarification expert model, risk avoidance decision expert model, equipment optimization expert model, and data assimilation and alignment expert model (including crowdsourced data assimilation and multimodal spatiotemporal alignment). The system includes a dialect recognition expert model, which takes a voice signal as input and outputs standard text after dialect decoding; a semantic clarification expert model, which is responsible for human-computer dialogue interaction, guiding users to clarify ambiguous semantics through proactive questioning; a disaster avoidance decision-making expert model, which generates escape routes based on BeiDou satellites, takes the flood and rainstorm warning level and the user's location coordinates as input, and outputs dynamic disaster avoidance dialogue strategies including evacuation directions; an equipment optimization expert model, which resolves the contradiction between limited equipment resources and the computing power requirements of AI (Artificial Intelligence) in disaster scenarios, ensuring battery life and performance, and ensuring that local computing power can still support critical operations even in the event of sudden cloud interruptions; and a data assimilation and alignment expert model, which, on the one hand, converts crowdsourced observations such as water level photos and subjective water depth descriptions uploaded by community users into flow / water level inputs after quality assessment, assimilates the hydrological model in real time, and on the other hand, is used for spatiotemporal alignment of multi-source data and crisis credibility assessment, eliminating the deviation caused by sensor data and voice descriptions. In this embodiment of the invention, the first preset threshold can be 40%, and the second preset threshold can be 20%.

[0062] Specifically, embodiments of the present invention can determine the dialect recognition expert model, semantic clarification expert model, risk avoidance decision-making expert model, equipment optimization expert model, and data assimilation and alignment expert model in the flood disaster expert network model based on the operating status of the portable edge device. For the dialect recognition expert model, structured pruning is used to preserve the key dialect phoneme recognition path. Based on the dynamic MoE loading strategy, the expert network is triggered according to the remaining power of the device: when the power is >40%, the dialect recognition expert model, semantic clarification expert model, risk avoidance decision-making expert model, equipment optimization expert model, and data assimilation and alignment expert model are loaded simultaneously; when the power is 20% < power ≤ 40%, other local networks except for the equipment optimization expert model are shut down, and cloud-based collaborative inference is invoked; when the power is ≤20%, only the dialect recognition expert model, semantic clarification expert model, and risk avoidance decision-making expert model are retained, wherein the semantic clarification expert model enables semantic compression, retaining only noun phrases and verbs.

[0063] In addition, the system can connect to portable edge devices (such as smart bracelets) via a communication unit. When the main device's battery level is detected to be ≤10%, the bracelet will automatically wake up and vibrate to remind the user to enter low power mode and charge, ensuring the system can continue to operate. A schematic diagram of the overall system structure is shown below. Figure 3 As shown.

[0064] In this embodiment of the invention, the system completes the preparation of core data structures and models through initialization, and dynamically configures the expert network according to the equipment status, providing a guarantee for subsequent emergency interaction for the elderly during floods.

[0065] In step S102, an expert network model is used to acquire the user's voice stream data in real time using a portable edge device when the user is in a flood disaster emergency interaction scenario. The voice stream data is then preprocessed to extract voice features that meet preset clarity conditions.

[0066] It is understood that the user in the embodiments of the present invention can be an elderly user; the voice feature that meets the preset clarity condition can be a clear voice feature.

[0067] In practical implementation, this embodiment of the invention can perform voice acquisition and preprocessing. When entering a flood disaster emergency interaction scenario, the system activates the voice acquisition module and acquires the voice stream data of elderly users in real time through the microphone of a portable edge device. To address background interference such as environmental noise that may exist in the voice signal of elderly users, the system first performs information preprocessing, including noise reduction filtering and audio calibration, thereby extracting clear voice features.

[0068] The embodiments of the present invention can realize adaptive interaction and dynamic allocation of edge device computing resources in flood scenarios, so as to protect the lives and property of the elderly in the face of extreme flood weather.

[0069] It should be noted that the preset clear conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0070] Optionally, in one embodiment of the present invention, after acquiring the user's voice stream data in real time based on the portable edge device, the method further includes: if a preset keyword is detected in the voice stream data, pausing non-critical background tasks and increasing the weight of resource allocation to a preset percentage; if a modifier in the voice stream data is detected to be greater than a preset threshold, calling a prefix tree algorithm to analyze the syntax in order to generate the core semantic components of the statement.

[0071] It is understood that the preset keywords in the embodiments of the present invention can be disaster keywords, the preset percentage can be 70%, and the preset threshold can be 30%.

[0072] In actual implementation, the equipment optimization expert model in this embodiment of the invention can monitor voice content in real time to determine the level of emergency. When disaster keywords such as "flood," "rainstorm," and "submersion" are detected in the user's voice, the equipment optimization module immediately triggers resource reallocation, suspends non-critical background tasks, and increases the CPU (Central Processing Unit) computing resource allocation weight to 70% to ensure that relevant computing is prioritized during emergency interactions. In addition, if more than 30% of the user's statement contains uncertain modifiers, such as frequent use of words like "probably" or "maybe," indicating that the user's description may have obstacles, the equipment optimization model will automatically call the prefix tree algorithm for syntactic analysis to extract the core semantic components of the statement.

[0073] In this embodiment of the invention, the system can obtain preprocessed speech signals, providing conditions for subsequent speech recognition and analysis.

[0074] It should be noted that the preset keywords, preset percentages, and preset thresholds can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0075] In step S103, speech features that meet preset clarity conditions are analyzed to predict and complete the interrupted speech content of the speech stream data, thereby generating a complete speech sequence.

[0076] In this embodiment of the invention, speech discontinuity reconstruction and real-time analysis can be performed. The preprocessed clear speech features are input into the prefix tree algorithm module for real-time analysis, thereby predicting and completing the interrupted speech content of the speech stream data and generating a complete speech sequence.

[0077] This invention provides a prefix tree algorithm for reconstructing speech streams based on the discontinuous speech characteristics of elderly people in edge devices. Specifically, it includes:

[0078] Prefix Tree Initialization Module: Initializes the prefix tree structure, including defining the tree node storage unit. Each node contains the current phoneme segment, frequency statistics, and a set of child nodes.

[0079] Voice Acquisition and Preprocessing Module: Collects voice stream data of elderly users in real time through edge devices, extracts intermittent voice signals, and performs signal preprocessing to remove environmental noise and device background interference, and extracts effective phoneme features.

[0080] Real-time speech matching module: Based on the preprocessed phoneme features, it performs comparison and matching step by step through the prefix tree, traversing layer by layer from the root node, and updating the statistical frequency of the phoneme nodes in real time to reflect the usage frequency of the phonemes.

[0081] Speech Interruption Prediction Module: In response to the slow speech rate and frequent speech interruptions unique to the elderly, when the speech interruption time exceeds the set threshold (the threshold is set to 3 seconds) or when there is a fuzzy phoneme recognition result, the prefix tree node backtracking mechanism is automatically triggered. The backtracking path selects the phoneme path with the highest frequency to achieve speech content prediction and completion.

[0082] Dialect Recognition Calibration Module: The dialect recognition expert module is used to recognize and verify the accuracy of the predicted and completed speech sequence. When the recognition confidence is lower than the preset threshold (confidence is set to 80%), the cross-modal alignment expert module is automatically called to perform secondary calibration in combination with environmental sensor data.

[0083] Interactive Response Module: The complete, calibrated, and confirmed speech stream is handed over to the semantic clarification expert module for subsequent interactive response, and the response results drive the risk avoidance decision-making module to formulate specific emergency strategies.

[0084] Through the modular design described above, this invention addresses the intermittent and ambiguous speech characteristics of the elderly population in flood disaster scenarios, achieving efficient and real-time speech stream reconstruction and improving the reliability and effectiveness of disaster emergency interaction systems.

[0085] It should be noted that the preset clear conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0086] Optionally, in one embodiment of the present invention, analyzing speech features that meet preset clarity conditions to predict and complete interrupted speech content in speech stream data and generate a complete speech sequence includes: when a long-term interruption or blurred phoneme segment is detected in the speech stream data, triggering a node backtracking mechanism using a prefix tree algorithm to count the phoneme path frequency of speech features; backtracking to select a phoneme sequence path based on the phoneme path frequency to predict and complete interrupted speech content based on the phoneme sequence path and generate a complete speech sequence.

[0087] In actual implementation, after initialization with the prefix tree, the system gradually maps the phoneme stream into the prefix tree structure, matching phoneme segments layer by layer starting from the root node and dynamically updating the frequency statistics of the response nodes. Through real-time speech matching, pauses and repetitions in the speech of elderly individuals can be effectively captured. When a long time interval (e.g., an interruption exceeding a preset threshold of 3 seconds) or a blurry, difficult-to-distinguish phoneme segment is detected in the speech, the prefix tree algorithm triggers a node backtracking mechanism. Based on the accumulated phoneme path frequency statistics, the most probable phoneme sequence path is backtracked to predict and complete the interrupted speech content, thereby reconstructing the complete speech sequence.

[0088] In this embodiment of the invention, the system can compensate for the information loss caused by the elderly’s slow speech or disjointed speech, thereby generating a relatively complete and coherent speech signal.

[0089] In step S104, the crowdsourced data of the user's location is uploaded and verified. The verified crowdsourced data is then assimilated into a preset flood forecasting model to generate corrected flood level and flow forecasts. Flood risk information is then generated based on the corrected flood level and flow forecasts.

[0090] It is understood that embodiments of the present invention can incorporate a community-interactive crowdsourced data upload module to fully leverage the role of public participation in rainstorm and flood early warning. This module is designed for elderly users, providing a user-friendly interface and interactive dialogue, enabling them to promptly upload disaster information from their local area, including on-site photos, subjective estimates of hydrological depth, and textual descriptions. The system can receive this real-time information from the community, providing supplementary data sources for flood forecasting, especially in areas where official observations are scarce. This invention establishes community-participatory data collection and assimilation as one of the core technological pathways for accurate flood early warning and intelligent disaster avoidance decision-making.

[0091] In actual implementation, embodiments of the present invention can upload crowdsourced data of the user's location and verify the crowdsourced data, so assimilate the verified crowdsourced data into a preset flood forecasting model, generate corrected flood level and flow forecasts, and generate flood risk information based on the corrected flood level and flow forecasts.

[0092] This invention can use crowdsourced data from community interactions to assimilate and dynamically correct flood models in real time, further protecting the lives and property of the elderly in the face of extreme flooding weather.

[0093] Optionally, in one embodiment of the present invention, uploading and verifying crowdsourced data of the user's location, assimilating the verified crowdsourced data into a preset flood forecasting model to generate corrected flood level and flow forecasts, and generating flood risk information based on the corrected flood level and flow forecasts, includes: uploading crowdsourced data of the user's location and preprocessing the crowdsourced data to generate preprocessed crowdsourced data; cross-validating the preprocessed crowdsourced data to generate verified crowdsourced data; assimilating the verified crowdsourced data into a preset flood forecasting model to generate corrected flood level and flow forecast data; identifying the magnitude of flood risk changes based on the verified flood level and flow forecast data, and adjusting the flood threshold and warning level according to the magnitude of flood risk changes to generate flood risk information.

[0094] In actual implementation, such as Figure 4 As shown, this embodiment of the invention can perform data reporting and receiving: elderly users can report the real-time flood situation at their location through the voice, text, or photo functions of the dialogue terminal. Users can describe the water as "water above their calves" or upload photos of the water accumulation at the scene. The community interaction module of the system receives the information and automatically records the upload time, user location, and other data.

[0095] Furthermore, embodiments of the present invention can perform data preprocessing and quality assessment: the system preprocesses and performs quality checks on the received user data. For text descriptions, key information such as qualitative water level height (e.g., "calf" corresponds to approximately 0.5 meters) is extracted through parsing; for photos, the built-in image analysis algorithm is used to estimate the water depth or affected area reflected in the image. Next, the system cross-validates data from different sources: if multiple users report water levels in adjacent areas, the data corroborates each other to increase credibility; if a piece of data significantly deviates from common sense or surrounding data, it is marked as an anomaly and ignored.

[0096] Furthermore, embodiments of the present invention can perform data conversion and assimilation processing: for observation data that has passed quality control, the system performs format conversion and unit conversion as needed. The data assimilation module of the flood forecasting model based on the EnKF algorithm (Ensemble Kalman Filter) is activated, including a prediction phase and an assimilation phase, to integrate the latest observation data into the model state. Through iterative calculations, corrected flood level and flow forecast results are generated, thereby significantly improving forecast accuracy.

[0097] (1) Prediction Phase: The analysis state set from the previous time step is advanced to the current time step using the model state evolution equation to obtain the prediction set state. The state evolution equation can be expressed as:

[0098]

[0099] in, express The predicted state vector at time t. f (•) denotes the model evolution operator. This represents process noise and characterizes model error. This indicates the analysis state at step k-1 (the previous step). This indicates that the value is a prediction obtained solely through model evolution before assimilation. This is the time step index (representing the k-th time step). This indicates that the quantity is the optimal estimate obtained after fusion of observations.

[0100] The corresponding model observation equation is:

[0101]

[0102] in, express The observed state vector at time t, This is the observation matrix, used to map the model state to the observation space (such as water level, flow rate). Indicates observation noise. The predicted state vector at time t is denoted by t.

[0103] (2) Assimilation stage: when acquired Actual observation of time Then, the predicted set state is corrected by updating the formula using Kalman filtering.

[0104] First, calculate the sample covariance matrix based on the prediction set:

[0105]

[0106] in, Represents the sample covariance matrix. N Indicates the number of members in the set. This represents the average value of the predicted states of the set. This represents the predicted state vector of the i-th member at time k. This indicates that the value is a prediction obtained solely through model evolution before assimilation. For time step index, z is the index of the set member.

[0107] Next, calculate the Kalman gain:

[0108]

[0109] in, R Represents the observation error covariance matrix. Indicates Kalman gain, Represents the sample covariance matrix. This represents the transpose of H (i.e., projecting the observation space information back into the model space during computation). This is the observation matrix.

[0110] Finally, the observation information is incorporated into each ensemble member using Kalman gain, and the update formula is as follows:

[0111] i =1,… N

[0112] in, This represents the analysis state vector updated at time k after assimilation observation, where the i-th set member corresponds to an independent update. This represents the predicted state vector of the same member obtained solely through model evolution before assimilation prediction at time k. Represents the Kalman gain matrix. express Actual observation at any time For the observation matrix, For time step index, For collection member indexes.

[0113] Thus, the time is obtained. k The analysis set of states is used. The above assimilation and update process minimizes the deviation between model predictions and measured values, thus correcting the model state. The updated set of states serves as the initial condition for the next forecast time, and the prediction-assimilation process is executed iteratively in this way. Through this data assimilation module, the flood forecasting system can promptly absorb actual observation information, correct forecast water levels and flows in real time, and effectively improve forecast accuracy.

[0114] Furthermore, embodiments of the present invention can perform forecast correction and decision adjustment: the updated forecast results are fed back to the risk avoidance decision expert model in real time, and the warning level is automatically updated through a dynamic threshold adjustment algorithm. Whenever the hydrological model generates a new water level or risk forecast after data assimilation, the system decision module will immediately compare the new and old forecasts, identify the magnitude of changes in flood risk, and adjust the thresholds and warning levels accordingly. If the assimilated model shows that the water level in a certain area is rising faster than originally predicted, and the expected water level exceeds the preset safety threshold, the system will raise the warning level for that area and implement more timely evacuation or risk avoidance measures; conversely, if the threshold is not exceeded, the current warning level will be maintained and monitoring will continue. For areas that are no longer in danger, the warning level will be lowered and the prompts will be adjusted accordingly. The risk avoidance decision expert model will also update the content in the dialogue with the user accordingly to ensure that elderly users receive guidance based on the latest scenario. The entire threshold update process is completed by the system in real time without manual intervention. With the help of intelligent edge computing architecture, the above model updates and threshold adjustments can be carried out efficiently at edge nodes close to the site, minimizing cloud transmission latency; even if the central network is damaged in the event of a disaster, local devices can autonomously complete threshold reassessment based on sensor data, ensuring the continuous reliability of early warning services.

[0115] Furthermore, this embodiment of the invention allows for information feedback and iterative improvement: the system promptly feeds back updated flood risk information and evacuation plans to users through a dialogue interface, and confirms whether further prompts are needed based on user feedback. Simultaneously, all new user observation data and corrected model states are stored in the system database for subsequent analysis and further model improvement. As more community data is uploaded, the above quality control-assimilation-correction process is repeated to gradually optimize flood forecast accuracy, forming a virtuous cycle combining public participation and intelligent early warning.

[0116] In step S105, based on flood risk information, the dialect speech in the complete speech sequence is decoded into standard text, and the confidence level of the standard text is output.

[0117] In practical implementation, this embodiment of the invention can perform dialect recognition and multi-source data alignment calibration. For the reconstructed speech sequence obtained in the above steps, a dialect recognition expert model is invoked for speech recognition or text writing, decoding the dialect speech into standard text and outputting the recognition confidence score. When the recognition confidence score is higher than a preset threshold (80%), it indicates that the speech content is reliable, and the process directly proceeds to the next step of semantic clarification interaction. When the recognition confidence score is lower than the preset threshold (80%), the speech content may contain ambiguity or errors, and the cross-modal alignment expert module will be automatically activated for secondary calibration. The cross-modal alignment expert will acquire multi-source sensor data of the current environment, such as water level sensor data at disaster sites, the user's geographical coordinates, and the flood warning level of the corresponding area, and perform spatiotemporal consistency checks on the speech information.

[0118] This invention provides a method for multi-source data spatiotemporal alignment and crisis credibility assessment. Credibility assessment includes temporal credibility, spatial credibility, and semantic credibility. Temporal credibility is used to calibrate time-series consistency, verifying whether user-system voice interaction and water level sensor data belong to the same event cycle when there may be a time difference. Spatial credibility is used to calibrate the mapping relationship of geographical locations, avoiding cross-regional forecast misjudgments by verifying the spatial relationship between the location described by the voice interactor and the sensor monitoring point. Semantic credibility, calculated based on the number of dialogue turns and semantic entropy, is used to clarify any ambiguous semantics that users may express.

[0119] In one embodiment of the present invention, the cross-modal alignment expert weights and fuses temporal, spatial, and semantic credibility according to preset weights to calculate a multi-source data credibility value. C。 The formula for calculating the credibility of standard text is:

[0120]

[0121] in, C This represents the confidence value of the spatiotemporal alignment of multi-source data in the system. Indicates the weight of the time dimension. Indicates the spatial dimension weight. Represents the semantic dimension weight. This indicates the credibility score corresponding to the time dimension. This represents the credibility score corresponding to the spatial dimension. This represents the credibility score corresponding to the semantic dimension. Indicates a time-dimensional index. Indicates spatial dimension index, Represents a semantic dimension index;

[0122] First, time reliability is calculated based on the difference between the voice interaction time and the sensor recording time to determine whether they are within the same event cycle. The formula for calculating the reliability score corresponding to the time dimension is as follows:

[0123]

[0124] in, This indicates the credibility score corresponding to the time dimension. This represents the absolute difference between the time of voice interaction and the time recorded by the sensor. This indicates the maximum permissible time difference under the warning level;

[0125] Next, spatial credibility is calculated. Based on the distance between the user's location and the locations of each monitoring point, as well as the corresponding abnormal water level conditions, the degree of consistency between the user's described location and the data monitored by the sensors is evaluated. The closer the user is to the sensor monitoring point, the higher the credibility, and vice versa. The formula for calculating the credibility score corresponding to the spatial dimension is:

[0126]

[0127] in, This represents the credibility score corresponding to the spatial dimension. Indicates the user's location and the first i The distance between monitoring points Indicates the first i The water level status of each monitoring point (0 indicates normal, 1 indicates abnormal). Indicates the number of monitoring points. Indicates the monitoring point index;

[0128] Finally, semantic credibility is calculated based on the number of dialogue rounds to quantify the clarity of the user's semantic description and the number of questions asked for clarification. n The more semantic dimensions there are, or the higher the semantic uncertainty, the lower the semantic credibility. The formula for calculating the credibility score corresponding to the semantic dimension is:

[0129]

[0130] in, This represents the credibility score corresponding to the semantic dimension. n This indicates the number of times a question is asked to clarify semantics; to avoid fatigue, this number is set to... n ≤5.

[0131] The mathematical constraints for weighting spatiotemporal credibility are:

[0132]

[0133] To improve the accuracy of credibility assessment, the system dynamically adjusts the weights of each dimension based on the environment, specifically including:

[0134] (1) When the user's positioning signal is weak (positioning error exceeds 50 meters), the weight of the spatial dimension is reduced to below 20%, and the importance of the time dimension is enhanced to compensate for the positioning uncertainty with time correlation;

[0135] (2) When the external sensor data update delay is large (more than 10 minutes), the weight of the time dimension is reduced to below 20%, and the weight of the spatial dimension is increased to make up for the lack of timeliness of the sensor data through spatial correlation.

[0136] (3) When the current weather warning level is red (highest alert), in order to avoid missing the opportunity due to over-reliance on semantic interaction, the weights of time, space and semantic dimensions can be set to 0.2 in a balanced manner, that is, more emphasis is placed on using regional monitoring data to assist decision-making, and the reliance on user semantic confirmation is reduced.

[0137] (4) When a user’s voice is detected to have obvious ambiguity (speech rate exceeding 180 words / minute or speech interruption exceeding 3 seconds), the semantic dimension weight is increased to 0.4, and the overall credibility assessment is improved by strengthening semantic analysis.

[0138] After the above multi-source data alignment and credibility calculation, the system obtains a comprehensive credibility value. C And adjust the emergency response strategy accordingly: if C If the value is high (e.g., not lower than 0.8), the system considers the user's voice to be a good match with the environmental conditions, and an emergency plan can be formulated directly based on this information; if... C If the value is too low, the system will rely more on objective sensor data or adopt a conservative strategy when generating risk avoidance strategies, and may need to proceed to the next step to interact with the user to obtain more information.

[0139] In step S106, based on the confidence level of the standard text, questions are posed to the user to guide them in clarifying ambiguous semantics, thereby generating user semantic information and environmental data that meet the preset high confidence level conditions.

[0140] It is understood that the user semantic information and environmental data that meet the preset high credibility conditions in the embodiments of the present invention can be user semantic information and environmental data with high credibility.

[0141] In practical implementation, this invention can perform semantic clarification interaction and semantic credibility enhancement. After determining the approximate textual expression of the speech content, the system enters the semantic clarification interaction stage. Based on the confidence level of the standard text, a semantic clarification expert model is used to conduct multiple rounds of dialogue with the user to guide the user to clarify ambiguous semantics, generating user semantic information and environmental data that meet preset high credibility conditions, i.e., eliminating possible ambiguities and vague semantics, and further improving semantic credibility. The system will proactively ask questions about the unclear parts of the content identified in the previous stage, guiding the user to clarify key information, such as location details, the degree of danger, or special needs. After each round of clarification inquiry, the system records the number of dialogue rounds n and calculates the current semantic entropy, comprehensively evaluating the trend of semantic credibility changes. After several rounds of question and answer (excluding fatigue considerations, no more than 5 rounds), the user's intent and description tend to be clear and explicit, the semantic entropy decreases significantly, and the semantic credibility increases accordingly. When the preset upper limit of dialogue rounds (e.g., 5 rounds) is reached or the system determines that the semantics are sufficiently clear, the semantic clarification stage ends. Through this interaction process, the system ensures that it understands the user's intent and the situation on site to the greatest extent possible, laying a semantic foundation for generating the optimal emergency avoidance strategy.

[0142] It should be noted that the preset high confidence conditions can be set by those skilled in the art according to the actual situation, and no specific restrictions are imposed here.

[0143] In step S107, based on user semantic information and environmental data that meet preset high credibility conditions, a dynamic risk avoidance dialogue strategy for rainstorms and floods is generated according to the flood and rainstorm warning level and the user's location coordinates.

[0144] Specifically, embodiments of the present invention can formulate and output risk avoidance decisions. After acquiring highly reliable user semantic information and environmental data, the system generates specific flood disaster emergency avoidance dialogue strategies using a risk avoidance decision-making expert model. The expert model combines authoritative early warning information and the user's current geographical location, utilizing BeiDou satellite positioning data to plan safe escape routes. Based on comprehensive multi-source information, a dynamic risk avoidance plan is formulated, including evacuation direction, route selection, and possible safe locations. For example, if the water level in the user's area is detected to be rising continuously and the warning level is high, the user can immediately move to higher ground along a predetermined route, with simultaneous voice or screen prompts guiding the evacuation direction. If the reliability assessment indicates a discrepancy between the user's reported danger and the sensor data, the system will adopt a conservative strategy, such as reminding the user to be vigilant and seek a safe area nearby for further confirmation. During the decision-making process, the equipment optimization expert model continues to ensure the system operates in a high-performance, low-energy-consumption manner, and can call upon cloud computing resources if necessary to obtain more refined route planning. Finally, the generated emergency avoidance strategy is fed back to the user through voice synthesis or terminal display.

[0145] Through the above steps, the system in this invention identifies and reconstructs the dialect and intermittent speech characteristics of elderly users, integrates multi-source environmental data to calibrate information credibility, optimizes equipment resources in real time, and outputs personalized risk avoidance decisions. It realizes intelligent interaction throughout the entire process from voice input to emergency strategy output, improving the reliability and timeliness of human-computer interaction in emergency situations such as floods.

[0146] Specifically, it can be combined with Figure 5 As shown, the working principle of the generative rainstorm and flood intelligent risk avoidance dialogue method in this embodiment of the invention is explained in detail with a specific example.

[0147] like Figure 5 As shown, embodiments of the present invention may include the following steps:

[0148] Step S501: Voice input for elderly users.

[0149] Step S502: Dialect recognition.

[0150] Step S503: Crowdsourced data assimilation.

[0151] Step S504: Semantic clarification.

[0152] Step S505: Crowdsource data credibility assessment (time, space, semantics).

[0153] Step S506: Risk avoidance decision generation.

[0154] According to the generative intelligent flood avoidance method proposed in this embodiment, the MoE architecture works collaboratively through multiple "expert networks." While ensuring model capacity, it activates only experts relevant to the current task, thereby improving computational efficiency. Based on this dynamic MoE architecture, a flood emergency interaction system for the elderly is designed, deploying expert networks for dialect recognition, semantic clarification, hazard decision-making, device optimization, and cross-modal alignment on edge devices. During interaction, the system dynamically calls upon experts as needed: it can balance local computing power and battery life through device-optimized experts, switching to cloud-based collaborative computing when necessary to ensure real-time response under peak interaction volumes; and it can utilize cross-modal alignment experts to fuse sensor data to verify the reliability of voice information, providing elderly users with timely and accurate hazard avoidance guidance. This addresses several issues: the high cost of frequent calls to existing large models, which could trigger millions of interaction requests daily during rainstorm disasters, potentially causing a "burst" of computing resources; the excessively long training period for dialect data in the Claude model, making it difficult to meet the emergency management requirement of "system readiness within 30 minutes before disaster"; the reliance on fixed monitoring stations for traditional flood forecasting, which has limited spatial coverage; and the current intelligent interactive system's failure to integrate the "community interaction-data assimilation-dynamic correction" closed loop to the edge, resulting in a disconnect between forecasts and actual flood conditions.

[0155] Next, referring to the accompanying drawings, we describe the generative rainstorm and flood intelligent risk avoidance dialogue system proposed according to an embodiment of the present invention.

[0156] Figure 6 This is a schematic diagram of the structure of the generative rainstorm and flood intelligent disaster avoidance dialogue system according to an embodiment of the present invention.

[0157] like Figure 6 As shown, the generative rainstorm and flood intelligent disaster avoidance dialogue system 10 includes: a prefix tree initialization module 100, a preprocessing module 200, a prediction module 300, a crowdsourced data upload module 400, a decoding module 500, an interaction module 600, and a disaster avoidance dialogue module 700.

[0158] Specifically, the prefix tree initialization module 100 is used to initialize the data structure of the prefix tree and load the expert network model under the dynamic hybrid expert architecture according to the operating status of the portable edge device.

[0159] The preprocessing module 200 is used to utilize an expert network model to acquire the user's voice stream data in real time using a portable edge device when the user is in a flood disaster emergency interaction scenario, and to preprocess the voice stream data to extract voice features that meet preset clarity conditions.

[0160] The prediction module 300 is used to analyze speech features that meet preset clarity conditions in order to predict and complete the interrupted speech content of the speech stream data and generate a complete speech sequence.

[0161] The crowdsourced data upload module 400 is used to upload crowdsourced data of the user's location and verify the crowdsourced data. The verified crowdsourced data is then assimilated into a preset flood forecasting model to generate corrected flood level and flow forecasts. Based on the corrected flood level and flow forecasts, flood risk information is generated.

[0162] The decoding module 500 is used to decode dialect speech in a complete speech sequence into standard text based on flood risk information, and output the confidence level of the standard text.

[0163] The interaction module 600 is used to ask questions to the user based on the confidence level of standard text, so as to guide the user to clarify ambiguous semantics and generate user semantic information and environmental data that meet the preset high confidence conditions.

[0164] The risk avoidance dialogue module 700 is used to generate dynamic risk avoidance dialogue strategies for rainstorms and floods based on user semantic information and environmental data that meet preset high credibility conditions, according to the flood and rainstorm warning level and the user's location coordinates.

[0165] Optionally, in one embodiment of the present invention, the prefix tree initialization module 100 includes: a determination unit, a first loading unit, a second loading unit, and a third loading unit.

[0166] The determining unit is used to determine the dialect recognition expert model, semantic clarification expert model, risk avoidance decision expert model, equipment optimization expert model, and data assimilation and alignment expert model in the flood disaster expert network model based on the operating status of the portable edge device.

[0167] The first loading unit is used to simultaneously load a dialect recognition expert model, a semantic clarification expert model, a risk avoidance decision expert model, a device optimization expert model, and a data assimilation and alignment expert model when the remaining power of the portable edge device is greater than a first preset threshold.

[0168] The second loading unit is used to load the device optimization expert model when the remaining power is greater than the second preset threshold and less than or equal to the first preset threshold.

[0169] The third loading unit is used to load the dialect recognition expert model, the semantic clarification expert model, and the risk avoidance decision expert model when the remaining power is less than or equal to the second preset threshold.

[0170] Optionally, in one embodiment of the present invention, the generative rainstorm and flood intelligent disaster avoidance dialogue system 10 further includes: a pause module and a call module.

[0171] The pause module is used to pause non-critical background tasks and increase the weight of resource allocation to a preset percentage when the voice stream data is detected to contain preset keywords after the user's voice stream data is acquired in real time by the portable edge device.

[0172] The calling module is used to call the prefix tree algorithm to analyze the syntax and generate the core semantic components of the sentence when the number of modifiers detected in the speech stream data exceeds a preset threshold.

[0173] Optionally, in one embodiment of the present invention, the prediction module 300 includes a statistical unit and a prediction unit.

[0174] The statistical unit is used to trigger a node backtracking mechanism using a prefix tree algorithm when long-term interruptions or ambiguous phoneme segments are detected in the speech stream data, in order to count the phoneme path frequency of speech features.

[0175] The prediction unit is used to backtrack and select the phoneme sequence path based on the phoneme path frequency, so as to predict and complete the interrupted speech content based on the phoneme sequence path and generate a complete speech sequence.

[0176] Optionally, in one embodiment of the present invention, the crowdsourced data upload module 400 includes: an upload unit, a cross-validation unit, an assimilation unit, and an adjustment unit.

[0177] The upload unit is used to upload crowdsourced data from the user's location and preprocess the crowdsourced data to generate preprocessed crowdsourced data.

[0178] The cross-validation unit is used to cross-validate the preprocessed crowdsourced data to generate validated crowdsourced data.

[0179] The assimilation unit is used to assimilate the validated crowdsourced data into a preset flood forecasting model to generate corrected flood level and flow forecast data.

[0180] The adjustment unit is used to identify the magnitude of changes in flood risk based on corrected flood level and flow forecast data, and adjust the flood threshold and warning level according to the magnitude of changes in flood risk to generate flood risk information.

[0181] Optionally, in one embodiment of the present invention, the formula for calculating the credibility of standard text is:

[0182]

[0183] in, C This represents the confidence value of the spatiotemporal alignment of multi-source data in the system. Indicates the weight of the time dimension. Indicates the spatial dimension weight. Represents the semantic dimension weight. This indicates the credibility score corresponding to the time dimension. This represents the credibility score corresponding to the spatial dimension. This represents the credibility score corresponding to the semantic dimension. Indicates a time-dimensional index. Indicates spatial dimension index, Represents a semantic dimension index;

[0184] The formula for calculating the credibility score corresponding to the time dimension is:

[0185]

[0186] in, This represents the absolute difference between the time of voice interaction and the time recorded by the sensor. This indicates the maximum permissible time difference under the warning level;

[0187] The formula for calculating the credibility score corresponding to the spatial dimension is:

[0188]

[0189] in, Indicates the user's location and the first i The distance between monitoring points Indicates the first i Water level status at each monitoring point Indicates the monitoring point index. Indicates the number of monitoring points;

[0190] The formula for calculating the credibility score corresponding to the semantic dimension is:

[0191]

[0192] in, n This indicates the number of times a question was asked to clarify the semantic meaning.

[0193] It should be noted that the foregoing explanation of the generative rainstorm and flood intelligent risk avoidance dialogue method embodiment also applies to the generative rainstorm and flood intelligent risk avoidance dialogue system of this embodiment, and will not be repeated here.

[0194] The generative intelligent dialogue system for rainstorm and flood avoidance proposed in this embodiment of the invention utilizes a MoE architecture that employs multiple "expert networks" to work collaboratively. While ensuring model capacity, it activates only experts relevant to the current task, thereby improving computational efficiency. Based on this dynamic MoE architecture, a flood emergency interaction system for the elderly is designed, deploying expert networks for dialect recognition, semantic clarification, hazard decision-making, device optimization, and cross-modal alignment on edge devices. During interaction, the system dynamically invokes experts as needed: it balances local computing power and battery life through device-optimized experts, switching to cloud-based collaborative computing when necessary to ensure real-time response under peak interaction volumes; and it leverages cross-modal alignment experts to fuse sensor data and verify the reliability of voice information, providing elderly users with timely and accurate hazard avoidance guidance. This addresses several issues: the high cost of frequent calls to existing large models, which could trigger millions of interaction requests daily during rainstorm disasters, potentially causing a "burst" of computing resources; the excessively long training period for dialect data in the Claude model, making it difficult to meet the emergency management requirement of "system readiness within 30 minutes before disaster"; the reliance on fixed monitoring stations for traditional flood forecasting, which has limited spatial coverage; and the current intelligent interactive system's failure to integrate the "community interaction-data assimilation-dynamic correction" closed loop to the edge, resulting in a disconnect between forecasts and actual flood conditions.

[0195] Figure 7 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. The electronic device may include:

[0196] The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0197] When the processor 702 executes the program, it implements the generative rainstorm and flood intelligent risk avoidance dialogue method provided in the above embodiments.

[0198] Furthermore, electronic devices also include:

[0199] Communication interface 703 is used for communication between memory 701 and processor 702.

[0200] The memory 701 is used to store computer programs that can run on the processor 702.

[0201] The memory 701 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0202] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0203] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0204] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0205] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described generative rainstorm and flood intelligent disaster avoidance dialogue method.

[0206] This invention also provides a computer program product storing a computer program that, when executed by a processor, implements the above-described generative rainstorm and flood intelligent disaster avoidance dialogue method.

[0207] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0208] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0209] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0210] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0211] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0212] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0213] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0214] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A generative intelligent dialogue method for rainstorm and flood disaster avoidance, characterized in that, Includes the following steps: Initialize the prefix tree data structure and load the expert network model under the dynamic hybrid expert architecture according to the operating status of the portable edge device; Using the expert network model, when a user is in an emergency interaction scenario during a flood disaster, the portable edge device acquires the user's voice stream data in real time and preprocesses the voice stream data to extract voice features that meet preset clarity conditions; Analyze the speech features that meet the preset clarity conditions to predict and complete the interrupted speech content of the speech stream data, and generate a complete speech sequence; Upload the crowdsourced data of the user's location and verify the crowdsourced data, so assimilate the verified crowdsourced data into the preset flood forecasting model, generate corrected flood level and flow forecasts, and generate flood risk information based on the corrected flood level and flow forecasts; Based on the flood risk information, the dialect speech in the complete speech sequence is decoded into standard text, and the confidence level of the standard text is output. Based on the confidence level of the standard text, questions are posed to the user to guide them in clarifying ambiguous semantics, thereby generating user semantic information and environmental data that meet preset high confidence conditions. Based on the user semantic information that meets the preset high credibility conditions and the environmental data, a dynamic flood and rainstorm avoidance dialogue strategy is generated according to the flood and rainstorm warning level and the user's location coordinates.

2. The generative intelligent dialogue method for rainstorm and flood disaster avoidance according to claim 1, characterized in that, The loading of the expert network model under the dynamic hybrid expert architecture based on the operating status of the portable edge device includes: Based on the operating status of the portable edge device, the following expert models are determined in the flood disaster expert network model: dialect recognition expert model, semantic clarification expert model, risk avoidance decision-making expert model, equipment optimization expert model, and data assimilation and alignment expert model. When the remaining battery power of the portable edge device is greater than a first preset threshold, the dialect recognition expert model, the semantic clarification expert model, the risk avoidance decision expert model, the device optimization expert model, and the data assimilation and alignment expert model are loaded simultaneously. If the remaining battery power is greater than the second preset threshold and less than or equal to the first preset threshold, the device optimization expert model is loaded. If the remaining battery power is less than or equal to the second preset threshold, the dialect recognition expert model, the semantic clarification expert model, and the risk avoidance decision expert model are loaded.

3. The generative intelligent dialogue method for rainstorm and flood disaster avoidance according to claim 1, characterized in that, After acquiring the user's voice stream data in real time using the portable edge device, the method further includes: If a preset keyword is detected in the voice stream data, non-critical background tasks are paused and the weight of resource allocation is increased to a preset percentage; If the number of modifiers in the speech stream data exceeds a preset threshold, the prefix tree algorithm is invoked to analyze the syntax in order to generate the core semantic components of the sentence.

4. The generative intelligent rainstorm and flood avoidance dialogue method according to claim 3, characterized in that, The analysis of speech features that meet preset clarity conditions to predict and complete interrupted speech content in the speech stream data, generating a complete speech sequence, includes: If a long-term interruption or ambiguous phoneme segment is detected in the speech stream data, the prefix tree algorithm is used to trigger a node backtracking mechanism to count the phoneme path frequency of the speech features. Phoneme sequence paths are selected by backtracking based on the phoneme path frequency, and the interrupted speech content is predicted and completed based on the phoneme sequence paths to generate the complete speech sequence.

5. The generative intelligent dialogue method for rainstorm and flood disaster avoidance according to claim 1, characterized in that, The process of uploading and verifying crowdsourced data for the user's location, assimilating the verified data into a preset flood forecasting model, generating corrected flood level and flow forecasts, and generating flood risk information based on the corrected flood level and flow forecasts includes: Upload the crowdsourced data of the user's location and preprocess the crowdsourced data to generate preprocessed crowdsourced data; Cross-validate the preprocessed crowdsourced data to generate the validated crowdsourced data; The verified crowdsourced data is assimilated into the preset flood forecast model to generate the corrected flood level and flow forecast data; Based on the corrected flood level and flow forecast data, the magnitude of flood risk changes is identified, and the flood threshold and warning level are adjusted according to the magnitude of flood risk changes to generate the flood risk information.

6. The generative intelligent dialogue method for rainstorm and flood disaster avoidance according to claim 1, characterized in that, The formula for calculating the credibility of the standard text is as follows: in, C This represents the confidence value of the spatiotemporal alignment of multi-source data in the system. Indicates the weight of the time dimension. Indicates the spatial dimension weight. Represents the semantic dimension weight. This indicates the credibility score corresponding to the time dimension. This represents the credibility score corresponding to the spatial dimension. This represents the credibility score corresponding to the semantic dimension. Indicates a time-dimensional index. Indicates spatial dimension index, Represents a semantic dimension index; The formula for calculating the credibility score corresponding to the time dimension is as follows: in, This represents the absolute difference between the time of voice interaction and the time recorded by the sensor. This indicates the maximum permissible time difference under the warning level; The formula for calculating the credibility score corresponding to the spatial dimension is as follows: in, Indicates the user's location and the first i The distance between monitoring points Indicates the first i Water level status at each monitoring point Indicates the monitoring point index. Indicates the number of monitoring points; The formula for calculating the credibility score corresponding to the semantic dimension is as follows: in, n This indicates the number of times a question was asked to clarify the semantic meaning.

7. A generative intelligent dialogue system for rainstorm and flood disaster avoidance, characterized in that, include: The prefix tree initialization module is used to initialize the prefix tree data structure and load the expert network model under the dynamic hybrid expert architecture according to the operating status of the portable edge device. The preprocessing module is used to utilize the expert network model to acquire the user's voice stream data in real time using the portable edge device when the user is in a flood disaster emergency interaction scenario, and to preprocess the voice stream data to extract voice features that meet preset clarity conditions. The prediction module is used to analyze the speech features that meet the preset clarity conditions in order to predict and complete the interrupted speech content of the speech stream data and generate a complete speech sequence. The crowdsourced data upload module is used to upload crowdsourced data of the user's location and verify the crowdsourced data, so assimilate the verified crowdsourced data into a preset flood forecasting model, generate corrected flood level and flow forecasts, and generate flood risk information based on the corrected flood level and flow forecasts. The decoding module is used to decode the dialect speech in the complete speech sequence into standard text based on the flood risk information, and output the confidence level of the standard text; The interaction module is used to ask questions to the user based on the confidence level of the standard text, so as to guide the user to clarify ambiguous semantics and generate user semantic information and environmental data that meet the preset high confidence conditions; The risk avoidance dialogue module is used to generate a dynamic risk avoidance dialogue strategy for rainstorms and floods based on the user's semantic information that meets the preset high credibility conditions and the environmental data, according to the flood and rainstorm warning level and the user's location coordinates.

8. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the generative rainstorm and flood intelligent disaster avoidance dialogue method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the generative rainstorm and flood intelligent risk avoidance dialogue method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement the generative rainstorm and flood intelligent risk avoidance dialogue method as described in any one of claims 1-6.

Images