Weather forecast and early warning interaction method based on artificial intelligence and related equipment

Abstract

The invention provides a weather forecasting and early warning interaction method based on artificial intelligence and related equipment, and belongs to the technical field of weather information forecasting, and the method comprises the steps: responding to a user interaction request, automatically obtaining the real-time position information of a user through a front-end interface, and obtaining the user input information; associating the user input information with the real-time position information, and inputting the user input information and the real-time position information into an AI large model for semantic understanding and intention recognition to obtain an intention recognition result; based on a visual workflow platform, executing an information acquisition process which is matched with the intention recognition result and is used in the meteorological and disaster field, and generating natural language response information which conforms to the intention of the user according to the obtained information acquisition result; and returning the natural language response information to the user side. According to the invention, the weather forecast and early warning interaction efficiency based on artificial intelligence and the accuracy of obtaining weather information are improved.

Description

Technical Field

[0001] This invention relates to the field of meteorological information forecasting technology, specifically to an artificial intelligence-based meteorological forecasting and early warning interactive method and related equipment. Background Technology

[0002] Weather forecasting and disaster warning services are crucial information support for the public in daily life, enabling them to plan their travel, arrange agricultural production, and prevent disaster risks. Traditionally, these services relied primarily on regular broadcasts through mass media such as television and radio. With the widespread adoption of mobile internet technology, the acquisition of meteorological information has gradually shifted to digital platforms. Currently, mainstream meteorological services are primarily provided to users through dedicated weather forecast applications (APPs) or WeChat official accounts of relevant organizations. These digital services constitute the core of current weather forecasting and disaster warning services, their basic purpose being to provide users with fundamental meteorological data such as temperature, precipitation, and wind speed for designated areas, and to issue warning signals for severe weather when necessary.

[0003] In the process of developing this invention, the inventors discovered that existing digital services of this type have at least the following shortcomings in practical applications: First, regarding data timeliness, most services rely on periodic batch updates from backend meteorological data sources, with update cycles typically lasting hours or longer. This model results in users being unable to obtain the latest warning information in a timely manner during sudden, localized severe weather events (such as short-duration heavy rainfall, thunderstorms, and strong winds), leading to delays in information transmission. Second, regarding user interaction experience, existing services mostly use graphical interface menu selection or simple keyword text matching for interaction. Users need to manually select or input specific parameters such as city and date to query the weather, a cumbersome process. Furthermore, existing systems cannot effectively distinguish whether user input is a serious meteorological information query, a disaster warning concern, or ordinary casual conversation, resulting in a lack of personalized services, especially in emergency situations where efficient and accurate information interaction cannot be provided.

[0004] In summary, existing technologies for weather forecasting and disaster early warning services suffer from insufficient real-time updates, high human-computer interaction costs, and low levels of intelligence, which hinder the improvement of efficiency and user experience in meteorological public services. Summary of the Invention

[0005] In view of this, it is necessary to provide an artificial intelligence-based weather forecast and early warning interaction method and related equipment to solve the technical problems of insufficient real-time updates of weather forecast and disaster early warning service information, high human-computer interaction operation costs, and low efficiency and poor user experience of weather services caused by low intelligence in existing methods.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides an artificial intelligence-based weather forecasting and early warning interaction method, comprising: In response to user interaction requests, the system automatically obtains the user's real-time location information and user input information through the front-end interface. The user input information is associated with the real-time location information and input into the AI ​​big model for semantic understanding and intent recognition to obtain the intent recognition result; Based on a visual workflow platform, an information acquisition process for the meteorological and disaster fields is executed that matches the intent recognition results, and natural language response information that conforms to the user's intent is generated based on the obtained information acquisition results. The natural language response information is returned to the user.

[0007] In one possible implementation, obtaining user input information includes: Receive text information submitted by the user through the text input interface; And / or, receive voice information submitted by the user through the voice input component; When the received information is voice information, the voice recognition service is invoked to convert the voice information into text information.

[0008] In one possible implementation, associating the user input information with the real-time location information and inputting it into an AI big data model for semantic understanding and intent recognition to obtain the intent recognition result includes: The context containing the real-time location information is concatenated or vectorized and fused with the text formed by the user input information to form enhanced input information; The enhanced input information is input into the AI ​​big model, and the intent recognition result is output through the intent classifier built into the AI ​​big model. The intent recognition result includes at least one of the following: weather query intent, disaster warning query intent, defense guide query intent, and casual conversation intent.

[0009] In one possible implementation, when the intent recognition result is a weather query intent, the execution of an information acquisition process for the meteorological and disaster field that matches the intent recognition result includes: The target query area is determined based on the real-time location information or the specified area information parsed from the user input information; Call the application programming interface (API) that interfaces with external meteorological data services to request real-time structured meteorological data for the target query area.

[0010] In one possible implementation, when the intent recognition result is a defense guide query intent, the execution of an information acquisition process for the meteorological and disaster field that matches the intent recognition result includes: Obtain current context information, which includes at least one of the real-time location information, currently valid meteorological data, or early warning information; Using the aforementioned contextual information as search criteria, the AI ​​big data model performs semantic retrieval on the internally structured disaster prevention knowledge base to obtain relevant prevention guide texts.

[0011] In one possible implementation, after the step of returning the natural language response information to the user, the AI-based weather forecast and early warning interaction includes: Obtain real-time feedback information from users, wherein the real-time feedback information includes at least one of text feedback and voice feedback; Based on the real-time feedback information, it is determined whether the natural language response information is correctly understood by the user or whether it meets the user's needs, and a judgment result is obtained. If the judgment result is negative, the AI ​​big model is triggered to adjust the content of the natural language response information or generate explanatory supplementary information based on the real-time feedback information to obtain additional response information, and the additional response information is returned to the user terminal.

[0012] In one possible implementation, the process of executing an information acquisition flow for the meteorological and disaster fields that matches the intent recognition result based on a visual workflow platform, and generating natural language response information that conforms to the user's intent based on the obtained information acquisition result, includes: In the visualization workflow platform, multiple processing branches are predefined to correspond to different intent recognition results; Based on the intent recognition result, the current task is routed to the corresponding processing branch; In the processing branch, data retrieval, knowledge retrieval, and the generation of natural language response information by calling the AI ​​big model are executed sequentially according to a predefined logical order.

[0013] On the other hand, the present invention also provides an artificial intelligence-based weather forecast and early warning interactive device, comprising: The information acquisition module is used to respond to user interaction requests by automatically acquiring the user's real-time location information and user input information through the front-end interface; The intent recognition module is used to associate the user input information with the real-time location information and input it into the AI ​​big model for semantic understanding and intent recognition to obtain the intent recognition result; The response generation module is used to execute an information acquisition process for the meteorological and disaster fields that matches the intent recognition result based on a visual workflow platform, and generate natural language response information that conforms to the user's intent based on the obtained information acquisition result. The response feedback module is used to return the natural language response information to the user terminal.

[0014] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the artificial intelligence-based weather forecast and early warning interactive method described in any of the above implementations.

[0015] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps in the artificial intelligence-based weather forecast and early warning interactive method described in any of the above implementations.

[0016] The beneficial effects of this invention are as follows: The artificial intelligence-based weather forecast and early warning interaction method provided by this invention significantly improves the real-time performance and accuracy of information services by connecting multiple links such as location perception, intent understanding, professional data acquisition, and natural language generation into an automated process. At the same time, the process is scheduled through a visual workflow platform, which makes the entire system have good maintainability and scalability, and can flexibly adapt to the ever-changing needs of meteorological services, which is conducive to improving the efficiency of meteorological services and user experience. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of an embodiment of the artificial intelligence-based weather forecast and early warning interaction method provided by the present invention; Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S101; Figure 3 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 4 For the present invention Figure 1A schematic diagram of an embodiment of S103; Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of S103; Figure 6 A schematic flowchart of another embodiment of the artificial intelligence-based weather forecast and early warning interaction method provided by the present invention; Figure 7 For the present invention Figure 1 A schematic flowchart of an embodiment of S103; Figure 8 A schematic diagram of an embodiment of the artificial intelligence-based weather forecast and early warning interactive device provided by the present invention; Figure 9 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] This invention provides an interactive method, device, electronic device, and storage medium for weather forecasting and early warning based on artificial intelligence. The technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] Figure 1 The following is a schematic flowchart of an embodiment of the artificial intelligence-based weather forecasting and early warning interaction method provided by the present invention, as shown below. Figure 1 As shown, the AI-based weather forecasting and early warning interaction method includes: S101. In response to user interaction requests, automatically obtain the user's real-time location information and user input information through the front-end interface.

[0025] Specifically, when a user initiates an interaction through an app, mini-program, official account, or similar platform, the system immediately responds to the request and performs two tasks in parallel: first, it automatically and seamlessly obtains the latitude and longitude coordinates of the user's current location through an integrated front-end SDK (such as the WeChat SDK); second, it receives the user's input information. This step ensures that subsequent processing has both the context of "where the user is" and the core content of "what the user wants to ask" in a quick and easy manner.

[0026] It should be noted that step S101 ensures the real-time nature and accuracy of location data. Latitude and longitude data are securely transmitted to the backend server via HTTPS and used as default parameters for subsequent queries. If the user explicitly specifies a region (e.g., "Beijing weather"), the specified information is used first; otherwise, latitude and longitude are used for rapid location. This improves the system's convenience and location accuracy.

[0027] In one specific alternative implementation, the location is updated based on whether the user interacts via button or voice. For example, if a user wants to check the weather in the city where their relatives or friends live, they can quickly obtain the core information by including the city name in their input.

[0028] The user input information adopts a variety of forms to improve user experience and convenience, including but not limited to: text, voice, and manual selection. If it is voice, it will be converted into text first. If there is a means to select, a recommendation list can be generated based on the user's historical usage records and preferences, and provided to the user for selection.

[0029] S102. Associate the user input information with the real-time location information and input it into the AI ​​big model for semantic understanding and intent recognition to obtain the intent recognition result.

[0030] Specifically, after acquiring basic information, the process enters the semantic understanding and intent recognition stage. This embodiment does not analyze the user's input text in isolation, but rather combines real-time location information as a key context, deeply fusing it with the input text, and inputting both into an AI model for deep semantic understanding. Based on the analysis of the fused information, the AI ​​model accurately identifies the user's deeper intent, such as whether they want to check the weather, learn about disaster warnings, seek defense guidelines, or simply engage in casual conversation.

[0031] Among them, large AI models include, but are not limited to, Generative Pre-trained Transformer (GPT), Bidirectional Encoder Representations from Transformers (BERT), A Robustly Optimized BERT Pre-training Approach (RoBERTa), A Lite BERT (ALBERT), Text-to-Text Transfer Transformer (T5), Enhanced Representation through Knowledge Integration (ERNIE), and Chat Generative Pre-trained Transformer (ChatGPT).

[0032] S103. Based on a visual workflow platform, execute an information acquisition process for the meteorological and disaster fields that matches the intent recognition results, and generate natural language response information that conforms to the user's intent based on the obtained information acquisition results.

[0033] Specifically, this step, based on the clearly defined intent recognition results, invokes a visual workflow platform to execute highly targeted professional information acquisition and integration tasks. This visual workflow platform routes tasks to pre-defined, specific processing branches dedicated to the meteorological and disaster fields, according to different intent types. For example, for a weather query intent, the branch process calls an external meteorological data API to obtain real-time structured data for a specified area; for a warning query intent, it redirects to query an internal warning database. Each branch, following predefined logic, systematically completes the acquisition of accurate information from external or internal data sources, and ultimately, this structured information is transformed into a fluent, natural text response that conforms to the user's query context by an AI big data model.

[0034] Among them, visualization workload platforms include, but are not limited to: dedicated low-code / no-code AI application development platforms (such as Dify), process automation and business rule engines, cloud integration platform as a service (iPaaS), etc.

[0035] Preferably, this embodiment uses Dify to build a visual workload platform. Dify allows developers to assemble multiple nodes, including large model calls, knowledge base retrieval, condition judgments, and API calls, through a graphical interface. It is very suitable for building the complex process in this embodiment that routes to different processing branches according to different intentions.

[0036] S104. Return the natural language response information to the user.

[0037] Step S104 will return the natural language response information, which integrates real-time data and professional knowledge explanations, to the user through the original interaction channel.

[0038] To illustrate this process, consider the following example: A user located in the coastal city of Xiamen sends a voice message on the official WeChat account during typhoon season asking, "Is the wind strong now? What precautions should I take?" The system automatically retrieves the user's location information ("Xiamen City") and converts the voice message into text. The AI ​​model combines the location and text to accurately identify the combined intent of "weather status inquiry" and "typhoon prevention guidance consultation." Subsequently, the workflow platform initiates two parallel branches: one branch calls the weather API to obtain real-time wind speed and wind force level data for Xiamen; the other branch searches the prevention knowledge base based on the context of "Xiamen" and "typhoon." Finally, the AI ​​model integrates the information from both locations to generate a coherent response: "The coastal wind force in Xiamen is currently at level 7, with gusts of level 8-9. During the typhoon's impact, it is recommended that you stay away from the coast, reinforce doors and windows, and be aware of the risk of falling objects caused by strong winds." This embodiment connects multiple stages, including location awareness, intent understanding, professional data acquisition, and natural language generation, into an automated and visualized process, significantly improving the real-time performance and accuracy of information services. At the same time, the process is scheduled through a visualized workflow platform, giving the entire system good maintainability and scalability, enabling it to flexibly adapt to the ever-changing needs of meteorological services, and thus improving the efficiency of meteorological services and user experience.

[0039] In some embodiments of the present invention, such as Figure 2 As shown, step S101, obtaining user input information, includes: S201. Receive text information submitted by the user through the text input interface; S202, and / or receive voice information submitted by the user through the voice input component, wherein when the received information is voice information, the voice recognition service is invoked to convert the voice information into text information.

[0040] In this embodiment, the process of obtaining user input information is constructed as an access layer that supports multimodal interaction.

[0041] Specifically, this step involves receiving text information directly entered or pasted by the user through a text input interface deployed on the front end. This is the most basic and direct interaction method.

[0042] Meanwhile, the integrated voice input component allows users to submit voice information via audio acquisition devices such as microphones, thus broadening the interaction channels. When the system detects that the input signal is voice, it immediately invokes the integrated speech recognition service to process the voice information. This speech recognition service may be based on a pre-built speech recognition engine or utilize interfaces such as WeChat's voice recognition interface or third-party automatic speech recognition (ASR) cloud services. The processing typically includes preprocessing the raw audio signal (such as noise reduction and frame segmentation), feature extraction, and decoding via acoustic and language models, ultimately converting the continuous speech stream into an accurate text sequence. This achieves the transformation of unstructured speech information into standardized text that can be processed by subsequent large-scale AI models.

[0043] For example, while driving, a user can directly say the voice command "Check the rainfall situation in the current area" to their phone. The audio is captured by the voice input component, and then the voice recognition service converts it into the text message "Check the rainfall situation in the current area." This text message will then proceed through the semantic understanding process, just like text input obtained through other means.

[0044] It should be noted that text input is suitable for situations requiring precise expression or in noisy environments where speaking is inconvenient; while voice input frees up the user's hands and eyes, and is particularly optimized for interactive experiences in situations such as mobility, when hands are occupied (e.g., driving, working), or in accessible environments. Users do not need to navigate complex interfaces or manually input data. Secondly, by uniformly converting voice information into text information, the system provides a consistent and structured processing object for subsequent semantic understanding and intent recognition modules.

[0045] This embodiment processes content acquired through different input methods to obtain a consistent and structured processing object. This allows subsequent large-scale AI model processing to focus on text semantic analysis without having to establish two separate processing logics for speech and text. This simplifies the system architecture, reduces the coupling between various functional modules within the system, improves the consistency and processing efficiency of the overall process, and lowers the barrier for users to use intelligent weather services, thereby enhancing the user experience.

[0046] In some embodiments of the present invention, such as Figure 3 As shown, step S102 associates user input information with real-time location information and inputs it into the AI ​​big data model for semantic understanding and intent recognition, obtaining intent recognition results, including: S301. Concatenate or vectorize the text formed by the context containing real-time location information and the user input information to form enhanced input information; S302. Input the enhanced input information into the AI ​​big model, and output the intent recognition result through the intent classifier built into the AI ​​big model. The intent recognition result includes at least one of the following: weather query intent, disaster warning query intent, defense guide query intent, and casual conversation intent.

[0047] Specifically, the real-time location information obtained from the front end (such as latitude and longitude or its converted administrative division name) is first integrated as a structured context fragment with the preprocessed user input information (such as after speech recognition conversion).

[0048] In this embodiment, user input information integration can be achieved through direct text concatenation, such as appending a contextual identifier like "[Location: Haidian District, Beijing]" before the user's query text. Alternatively, a more complex vectorized fusion method can be used, where the location context and user text are converted into high-dimensional semantic vectors, and then fused through a specific neural network layer. This combines geographic location semantics with user query semantics in vector space, forming a richer and more comprehensive enhanced input. This step ensures that subsequent intelligent analysis is based on a complete context of "who, where, and what was asked."

[0049] After the enhanced input information is fed into the AI ​​large model (such as Qwen-plus), the built-in intent classifier of the model performs the recognition task.

[0050] This intent classifier, trained on a large dataset of dialogues labeled with categories such as "weather query", "disaster warning query", "defense guide query", "casual chat", and "sensitive content", is able to accurately determine the deeper purpose of user input.

[0051] For example, when the augmented input is "[Location: Shenzhen] What's the weather like?", the model can combine the location context of "Shenzhen" to more accurately classify it as a "weather query intent" rather than a general "casual chat". Similarly, for the input "Do I need to prepare for the wind here?", the model, combined with specific location information, can determine that it is more likely to be a "defense guide query intent" rather than a simple "weather query". This process not only completes the intent classification but also provides clear guidance for subsequent processes.

[0052] Take a specific scenario as an example: A user located in Shanghai sends a voice message, "How big is the impact of the typhoon?" The system converts this message into text through speech recognition and integrates it with automatically acquired location information for "Shanghai" to form enhanced input. After analyzing this enhanced input, the AI ​​model accurately outputs "disaster warning query intent." Conversely, if the same user inputs "My mood is as good as the weather today," even with location information, the model can accurately identify it as "casual conversation intent," thus triggering a corresponding friendly response mechanism instead of initiating an unnecessary data query process.

[0053] This embodiment effectively eliminates ambiguity in user input caused by vague regional references or lack of context by introducing location context, making intent judgment more accurate and reliable. This is especially crucial for meteorological and disaster early warning services that heavily rely on geographical location, and provides a key basis for distinguishing between professional queries and generalized casual conversation. By combining location information, it better distinguishes whether a user is concerned about the weather conditions in a specific area or simply expressing a weather-related emotion, thus achieving a balance between professional service response and a humanized interactive experience. This ensures the consistency and efficiency of the entire intelligent interaction chain from understanding to execution, and improves the overall service quality and user satisfaction of the meteorological early warning intelligent agent.

[0054] In some embodiments of the present invention, such as Figure 4 As shown, in step S103, when the intent recognition result is a weather query intent, an information acquisition process for the meteorological and disaster fields that matches the intent recognition result is executed, including: S401. Determine the target query area based on real-time location information or specified area information parsed from user input information; S402. Call the application programming interface that interfaces with the external meteorological data service to request real-time structured meteorological data for the target query area.

[0055] Specifically, when the user's intent is determined to be a weather query, an automated process specifically designed to acquire timely meteorological data is initiated. To achieve this, this embodiment employs a fusion-based regional determination method.

[0056] Specifically, the system first attempts to use a large AI model to perform named entity recognition from the user's natural language input, directly parsing out the explicitly specified region name. For example, if a user asks, "What's the temperature in Shanghai today?", "Shanghai" can be directly identified as the target region. If the user's input does not contain a specific place name, such as simply saying, "Will it rain today?", the system will automatically fall back to and rely on the user's real-time latitude and longitude coordinates obtained by the front-end interface (such as the WeChat SDK) at the initial stage of the interaction. By combining this coordinate information with geocoding services (such as the Tencent Maps API), it can be converted into corresponding administrative division information such as province, city, and district, thus seamlessly determining the "target query area." This mechanism ensures that the system can accurately understand the user's geographic location intent regardless of how the user expresses it.

[0057] After identifying the target query area, the information acquisition process enters the data acquisition phase. This embodiment does not use static or cached forecast data, but instead initiates a real-time data request through a pre-configured application programming interface (API) that interfaces with a professional third-party meteorological data service platform (such as the "Hefeng Weather" API). This request sends the identified regional information as key parameters to the external meteorological service in the form of a structured query. Upon receiving the request, the external service returns the latest structured meteorological data package for that region, typically covering multiple indicators such as temperature, humidity, air pressure, wind speed, wind direction, precipitation, UV index, and detailed forecasts for the next few hours. This data package is returned in a machine-readable format (such as JSON), providing an accurate and real-time data source for subsequent generation of natural language responses.

[0058] This embodiment effectively overcomes the information delay problem caused by data caching or timed updates in traditional applications by directly connecting to the real-time interface of professional meteorological data sources, ensuring that the meteorological information returned to users has high timeliness and accuracy. At the same time, the integration of location determination and data acquisition also significantly improves response speed.

[0059] In some embodiments of the present invention, such as Figure 5 As shown, in step S103, when the intent recognition result is a defense guide query intent, an information acquisition process for the meteorological and disaster fields that matches the intent recognition result is executed, including: S501. Obtain current context information, which includes at least one of real-time location information, currently valid meteorological data, or early warning information. S502. Using contextual information as the retrieval criteria, semantic retrieval is performed on the internally structured disaster prevention knowledge base through an AI big data model to obtain relevant prevention guide texts.

[0060] When a user's intent is identified as a defense guide query, the information retrieval process executed in this embodiment does not return static, generic text, but instead initiates an intelligent retrieval and generation process based on deep context awareness.

[0061] Specifically, the first step is to obtain multi-dimensional contextual information related to the current dialogue. This multi-dimensional contextual information includes at least: real-time location information automatically obtained by the front end or specified by the user (e.g., the user's city or the specific area being queried); real-time meteorological data currently valid in the area obtained through previous steps or parallel queries (e.g., temperature, precipitation probability, wind speed and direction); and structured early warning information that may be in effect for the area (e.g., "orange rainstorm warning").

[0062] After acquiring the aforementioned contextual information that integrates spatiotemporal and event features, the AI ​​model is driven to perform semantic retrieval on an internally structured disaster prevention knowledge base. This knowledge base is not a simple collection of question-and-answer pairs, but rather a textual repository of prevention guidelines organized and stored in a structured manner, covering various disaster types and response scenarios. The retrieval process is not based on simple keyword matching, but rather on the AI ​​model performing deep semantic understanding and relevance calculations on the search criteria and knowledge base entries.

[0063] For example, when the context information is "Location: Shanghai; Effective Warning: Typhoon Blue Warning," the AI ​​model understands that the user's core need is to obtain "specific defensive measures for Shanghai in response to the typhoon blue warning." Subsequently, it semantically searches the knowledge base for content highly related to "typhoon defense," "coastal cities," and "blue warning level response," rather than simply entries containing the words "Shanghai" or "typhoon." The context and knowledge base content are converted into semantic vectors, and similarity matching and reordering are performed in the vector space to accurately locate the most relevant guide text fragments.

[0064] In a specific example, a user asks, "What should I do when a typhoon hits?", with the context including the user being located in "Ningbo City" and a "yellow typhoon warning" being issued locally. Instead of providing a nationwide typhoon preparedness manual, the system uses a large AI model to precisely retrieve and integrate targeted suggestions from a knowledge base based on the combined context of "Ningbo" (which may relate to the characteristics of a coastal city and key flood control areas) and "yellow warning" (corresponding to a specific level of defense response). For example, the system might suggest, "Ningbo City is currently under a yellow typhoon warning. You should immediately check and reinforce structures that are easily blown by the wind, move potted plants from your balcony indoors, and pay attention to local meteorological department notices regarding work stoppages, school closures, and business closures."

[0065] This embodiment dynamically combines and outputs the most targeted action suggestions based on the user's specific geographical location, real-time weather conditions, and warning levels, greatly enhancing the practical value and guiding significance of the information. At the same time, by introducing a large AI model for semantic retrieval, it can more accurately understand the user's complex query intent and rich contextual associations. Even if the knowledge base entries do not explicitly contain all contextual keywords, relevant content can be found through semantic associations, thereby improving the recall rate and accuracy of the retrieval.

[0066] In some embodiments of the present invention, when the intent recognition result is a disaster warning query intent, the execution of an information acquisition process for the meteorological and disaster fields that matches the intent recognition result includes: The target query area is determined based on real-time location information or specified area information parsed from user input. Access the internal early warning database to query the structured early warning information records currently in effect for the target query area.

[0067] Specifically, this embodiment determines a unique and specific "target query area" based on associated real-time location information (such as the administrative divisions corresponding to latitude and longitude automatically obtained by the front end) or explicit geographical indications parsed from the user's natural language input (such as "Beijing" or "Huangpu River shoreline"). This step ensures that subsequent queries have clear spatial directionality.

[0068] After identifying the target query area, the internal early warning database is accessed. This database is not a static data archive, but a continuously updated system used to record and manage various structured early warning information issued by the meteorological department. In this embodiment, a query is initiated into this database to request all early warning records for the aforementioned "target query area" that have not been lifted or expired. These early warning records are typically stored in a structured format, containing key fields such as early warning type (e.g., rainstorm, typhoon), early warning level (e.g., yellow, orange), issuance time, affected area, and effective period.

[0069] Compared to traditional services where users passively wait for push notifications or have to sift through complex text and image information, this embodiment allows users to directly inquire about the disaster risk of a specific area using natural language, and quickly locate and return accurate results from authoritative data sources, significantly improving the efficiency and accuracy of information acquisition.

[0070] In some embodiments of the present invention, such as Figure 6 As shown, after the step of returning the natural language response information to the user, this AI-based weather forecast and early warning interaction method includes: S601. Obtain real-time feedback information from users, including at least one of text feedback and voice feedback; S602. Based on real-time feedback information, determine whether the natural language response information is correctly understood by the user or whether it meets the user's needs, and obtain the judgment result. S603. If the judgment result is negative, the AI ​​big model is triggered to adjust the content of the natural language response information or generate explanatory supplementary information based on the real-time feedback information, obtain the supplementary response information, and return the supplementary response information to the user terminal.

[0071] To improve the accuracy of interaction and user satisfaction, this embodiment introduces a feedback and optimization mechanism after returning the initial natural language response information to the user. This feedback is not limited to explicit replies in the traditional form of text or voice, such as follow-up questions sent directly by the user like "I didn't understand" or "Please explain the specific impact range of the rainstorm warning again."

[0072] Specifically, after acquiring real-time multimodal feedback, the system analyzes and judges it to assess whether the previously returned natural language response information was correctly understood by the user and whether its content fully meets the user's deeper needs. For example, the system may analyze keywords in the text and the confidence level of emotional polarity in the speech to comprehensively determine whether the user is in a state of "understanding and satisfied," "confused," or "insufficient information." This judgment process is based on preset rule thresholds or is completed through a lightweight machine learning classification model.

[0073] If the assessment indicates that the initial response was not correctly understood or did not meet the requirements, an optimization process will be automatically triggered. At this point, the AI ​​big data model is invoked again, its input including not only the original dialogue context and query intent, but more importantly, the integrated real-time feedback information obtained this time. Based on this enhanced context, the AI ​​big data model adjusts and refines the content of the initial response information, or generates explanatory supplementary information. For example, if a user replies to a forecast of rain tomorrow with the text "Will it affect flights?", the system determines that the initial response lacks sufficient granularity. Subsequently, the AI ​​big data model can generate additional response information, such as "Rain is expected to start in the afternoon, which may cause delays to flights departing and arriving in the evening. We suggest you pay attention to the latest announcements from your airline," and return this additional information to the user.

[0074] This embodiment constructs a continuously perceptive and dynamically optimized intelligent interaction loop. By integrating multi-dimensional signals such as text and voice emotion for comprehensive judgment, it improves the accuracy of judging the user's implicit intentions and true satisfaction. It can make up for possible interpretation gaps or misunderstandings in real time during the interaction process, thereby effectively improving the reliability of meteorological warning information transmission and the overall intelligence level of the service.

[0075] In some embodiments of the present invention, such as Figure 7 As shown, step S103, based on a visual workflow platform, executes an information acquisition process for the meteorological and disaster fields that matches the intent recognition results, and generates natural language response information that conforms to the user's intent based on the obtained information acquisition results, including: S701. Predefine multiple processing branches corresponding to different intent recognition results in the visual workflow platform; S702. Based on the intent recognition result, route the current task to the corresponding processing branch; S703. In the processing branch, data retrieval, knowledge retrieval, and the generation of natural language response information by calling the AI ​​big model are executed in a predefined logical order.

[0076] The operation of this visual workflow platform includes the following steps: First, the visual workflow platform pre-configures several independent processing branches, each explicitly associated with a specific user intent recognition result (e.g., weather query, disaster warning query, defense guide query). These branches are constructed using visual logical nodes and connecting lines, defining a complete task sequence from receiving a specific intent to the final output response.

[0077] Once the user's input intent is recognized by the AI ​​model, this intent becomes a clear instruction signal input to the visualization workflow platform. The platform's routing mechanism then automatically directs the current processing task to a pre-defined branch that matches the intent. For example, if the user's intent is to query typhoon warnings for Shanghai, the task will be routed to a branch specifically for processing disaster warning queries.

[0078] Once the corresponding processing branch is entered, the system will strictly follow the predefined, sequential logical steps within that branch to perform the operation.

[0079] This embodiment transforms complex business logic into a flexibly configurable and visually manageable process, greatly improving the system's maintainability and scalability. When a new query type (such as air quality query) needs to be added or an existing process needs to be adjusted, configuration can be done through a graphical interface without requiring in-depth modification of the underlying code. This ensures that various query tasks can efficiently and accurately call the corresponding data sources and services according to their domain characteristics, effectively improving the efficiency, content quality, and accuracy of meteorological early warning services.

[0080] In some embodiments of the present invention, the AI-based weather forecasting and early warning interaction method further includes a step of supporting multi-turn dialogue: During a single session, the conversation state is maintained and updated using the AI ​​big model or associated caching mechanism. If a pronoun or omitted information referring to previous dialogue content is detected in subsequent user input, the user input is completed and understood based on the dialogue state, and the subsequent intent recognition and information acquisition process is executed based on the completed information.

[0081] Specifically, to simulate the continuity and contextual relevance of interpersonal dialogue, this embodiment adopts a conversational approach that supports multi-turn dialogues. This approach continuously maintains and dynamically updates the dialogue state. Specifically, during a user's continuous conversation, the key contextual information from the dialogue process is structured and saved using the short-term memory capabilities of the AI ​​model itself, or by associating it with an independent caching storage system (such as Redis), forming the "dialogue state." This information typically includes, but is not limited to: confirmed user location information, the most recently queried meteorological entity (such as city, disaster type), provided warnings or weather data conclusions, and subsequent concerns that the user may explicitly or implicitly express.

[0082] In subsequent interaction rounds, when new user input is received, the system first analyzes whether there are any pronouns or omissions that require the aforementioned context for accurate understanding. For example, after receiving a response to "What's the weather like in Beijing today?" in the first round, a user might only input "What about tomorrow?" or "What about the warning?". Here, "What about tomorrow?" refers to "the weather in Beijing," while "What about the warning?" might refer to "disaster warnings in the Beijing area." By detecting such pronouns (such as "that" or "it") or obvious omissions, and immediately retrieving the corresponding previous information (such as the location "Beijing," the topic "weather," or "warning") from the maintained "dialogue state," the system semantically completes the current brief and ambiguous user input, restoring its full intent (e.g., completed as "What's the weather like in Beijing tomorrow?" or "What are the disaster warnings for the Beijing area?").

[0083] The completed information, representing the user's complete statement, is then sent to the subsequent semantic understanding and intent recognition modules for processing, and the corresponding information retrieval process continues. This ensures that the dialogue can proceed smoothly around the same or related topics, without requiring the user to repeatedly state information already provided.

[0084] This embodiment effectively avoids intent recognition errors or query failures caused by information omissions through intelligent context completion, ensuring the accurate execution of complex and continuous meteorological consultation tasks, thereby significantly enhancing users' reliance on and satisfaction with intelligent meteorological early warning services.

[0085] In some embodiments of the present invention, the AI-based weather forecast and early warning interaction method serves multiple user requests through a concurrent processing mechanism: Maintain a thread pool in the backend service layer to receive concurrent user interaction requests; Each user interaction request is assigned an independent session thread, in which the complete process from obtaining user input information to returning natural language response information is executed independently.

[0086] To address the potential for sudden, high-concurrency access demands during meteorological disasters, this embodiment deploys a highly efficient concurrency processing mechanism at the backend service layer. This mechanism employs thread pool technology to manage and schedule concurrent user interaction requests.

[0087] Specifically, at startup or during operation, a thread pool with a pre-configured initial and maximum number of threads is created and maintained within the backend service container. When a large number of users initiate interactive requests through the frontend almost simultaneously, these requests are not processed sequentially but are received by the thread pool and added to its pending queue.

[0088] As a resource scheduler, the thread pool dynamically allocates or wakes up an independent session thread from its managed thread resources for each received user interaction request. Each session thread is a self-contained execution unit that independently carries and executes the complete and continuous processing flow from receiving input information from that specific user to finally generating and returning natural language response information. This includes a series of steps such as intent recognition, accessing the database or external API through the workflow platform, and calling the AI ​​model to generate the response. The various session threads are isolated from each other and do not interfere with each other, allowing complex weather data queries for user A and simple early warning inquiries for user B to be performed simultaneously without conflict.

[0089] This embodiment effectively avoids the huge overhead of frequently creating and destroying threads for each new request by reusing pre-created threads and processing requests in parallel. It also prevents system resource exhaustion and crashes caused by request backlog, thus ensuring uninterrupted service during critical moments such as disaster warnings. At the same time, by isolating each user session in an independent thread, it ensures the security and isolation of user data and processing state, avoiding data pollution or logical interference between different user requests.

[0090] To better implement the AI-based weather forecasting and early warning interaction method in this invention embodiment, based on the AI-based weather forecasting and early warning interaction method, correspondingly, as follows: Figure 8 As shown, this embodiment of the invention also provides an artificial intelligence-based weather forecast and early warning interactive device. The artificial intelligence-based weather forecast and early warning interactive device 800 includes: The information acquisition module 801 is used to respond to user interaction requests, automatically acquire the user's real-time location information through the front-end interface, and acquire user input information; The intent recognition module 802 is used to associate user input information with real-time location information and input it into the AI ​​big model for semantic understanding and intent recognition to obtain intent recognition results; The response generation module 803 is used to execute an information acquisition process for the meteorological and disaster fields that matches the intent recognition results based on a visual workflow platform, and generate natural language response information that conforms to the user's intent based on the obtained information acquisition results. The response feedback module 804 is used to return natural language response information to the user terminal.

[0091] The AI-based weather forecast and early warning interactive device 800 provided in the above embodiments can realize the technical solutions described in the above AI-based weather forecast and early warning interactive method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above AI-based weather forecast and early warning interactive method embodiments, which will not be repeated here.

[0092] like Figure 9 As shown, the present invention also provides an electronic device 900. The electronic device 900 includes a processor 901, a memory 902, and a display 903. Figure 9 Only some components of the electronic device 900 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0093] In some embodiments, processor 901 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 902 or process data, such as the artificial intelligence-based weather forecast and early warning interactive method of the present invention.

[0094] In some embodiments, processor 901 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 901 may be local or remote. In some embodiments, processor 901 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0095] In some embodiments, memory 902 may be an internal storage unit of electronic device 900, such as a hard disk or memory of electronic device 900. In other embodiments, memory 902 may also be an external storage device of electronic device 900, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 900.

[0096] Furthermore, the memory 902 may include both internal storage units of the electronic device 900 and external storage devices. The memory 902 is used to store application software and various types of data installed on the electronic device 900.

[0097] In some embodiments, display 903 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 903 is used to display information from electronic device 900 and to display a visual user interface. Components 901-903 of electronic device 900 communicate with each other via a system bus.

[0098] In one embodiment, when the processor 901 executes the AI-based weather forecast and early warning interactive program stored in the memory 902, the following steps can be implemented: In response to user interaction requests, the system automatically obtains the user's real-time location information and user input information through the front-end interface. The user input information is associated with real-time location information and fed into a large AI model for semantic understanding and intent recognition to obtain intent recognition results; Based on a visual workflow platform, an information acquisition process for the meteorological and disaster fields is executed that matches the intent recognition results, and natural language response information that conforms to the user's intent is generated based on the obtained information acquisition results. Return natural language response information to the user.

[0099] It should be understood that when the processor 901 executes the AI-based weather forecast and early warning interactive program in the memory 902, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0100] Furthermore, the embodiments of the present invention do not specifically limit the type of the electronic device 900 mentioned. The electronic device 900 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 900 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0101] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions of the artificial intelligence-based weather forecast and early warning interactive method provided in the above-described method embodiments.

[0102] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0103] The above provides a detailed description of the artificial intelligence-based weather forecasting and early warning interactive method, device, electronic device, and storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An interactive method for weather forecasting and early warning based on artificial intelligence, characterized in that, include: In response to user interaction requests, the system automatically obtains the user's real-time location information and user input information through the front-end interface. The user input information is associated with the real-time location information and input into the AI ​​big model for semantic understanding and intent recognition to obtain the intent recognition result; Based on a visual workflow platform, an information acquisition process for the meteorological and disaster fields is executed that matches the intent recognition results, and natural language response information that conforms to the user's intent is generated based on the obtained information acquisition results. The natural language response information is returned to the user.

2. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 1, characterized in that, The acquisition of user input information includes: Receive text information submitted by the user through the text input interface; And / or, receive voice information submitted by the user through the voice input component; When the received information is voice information, the voice recognition service is invoked to convert the voice information into text information.

3. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 1, characterized in that, The step of associating the user input information with the real-time location information and inputting it into an AI big data model for semantic understanding and intent recognition to obtain intent recognition results includes: The context containing the real-time location information is concatenated or vectorized and fused with the text formed by the user input information to form enhanced input information; The enhanced input information is input into the AI ​​big model, and the intent recognition result is output through the intent classifier built into the AI ​​big model. The intent recognition result includes at least one of the following: weather query intent, disaster warning query intent, defense guide query intent, and casual conversation intent.

4. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 3, characterized in that, When the intent recognition result is a weather query intent, the execution of an information acquisition process for the meteorological and disaster fields that matches the intent recognition result includes: The target query area is determined based on the real-time location information or the specified area information parsed from the user input information; Call the application programming interface (API) that interfaces with external meteorological data services to request real-time structured meteorological data for the target query area.

5. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 3, characterized in that, When the intent recognition result is a defense guide query intent, the execution of an information acquisition process for the meteorological and disaster field that matches the intent recognition result includes: Obtain current context information, which includes at least one of the real-time location information, currently valid meteorological data, or early warning information; Using the aforementioned contextual information as search criteria, the AI ​​big data model performs semantic retrieval on the internally structured disaster prevention knowledge base to obtain relevant prevention guide texts.

6. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 1, characterized in that, After returning the natural language response information to the user, the AI-based weather forecast and early warning interaction method includes: Obtain real-time feedback information from users, wherein the real-time feedback information includes at least one of text feedback and voice feedback; Based on the real-time feedback information, it is determined whether the natural language response information is correctly understood by the user or whether it meets the user's needs, and a judgment result is obtained. If the judgment result is negative, the AI ​​big model is triggered to adjust the content of the natural language response information or generate explanatory supplementary information based on the real-time feedback information to obtain additional response information, and the additional response information is returned to the user terminal.

7. The artificial intelligence-based weather forecasting and early warning interactive method according to claim 1, characterized in that, The aforementioned information acquisition process for the meteorological and disaster fields, based on a visual workflow platform, is executed in accordance with the intent recognition results, and natural language response information matching the user's intent is generated based on the obtained information acquisition results, including: In the visualization workflow platform, multiple processing branches are predefined to correspond to different intent recognition results; Based on the intent recognition result, the current task is routed to the corresponding processing branch; In the processing branch, data retrieval, knowledge retrieval, and the generation of natural language response information by calling the AI ​​big model are executed sequentially according to a predefined logical order.

8. An interactive weather forecasting and early warning device based on artificial intelligence, characterized in that, include: The information acquisition module is used to respond to user interaction requests by automatically acquiring the user's real-time location information and user input information through the front-end interface; The intent recognition module is used to associate the user input information with the real-time location information and input it into the AI ​​big model for semantic understanding and intent recognition to obtain the intent recognition result; The response generation module is used to execute an information acquisition process for the meteorological and disaster fields that matches the intent recognition result based on a visual workflow platform, and generate natural language response information that conforms to the user's intent based on the obtained information acquisition result. The response feedback module is used to return the natural language response information to the user terminal.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the artificial intelligence-based weather forecast and early warning interactive method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the artificial intelligence-based weather forecast and early warning interactive method described in any one of claims 1 to 7.

Images