Data Management Method and System for an Artificial Intelligence-Based Meteorological Service Platform

Abstract

This application relates to the field of data processing technologies, and in particular provides a method and system for data management of a meteorological service platform based on artificial intelligence. By deeply mining the semantic content of interaction information, the embodiments of this application can more precisely extract meteorological elements, and determine the relationship state evaluation weights based on the dense vectors of meteorological elements, so as to obtain a more accurate discriminant training error for meteorological content elements. This innovative method not only solves the deficiencies of traditional methods in processing unstructured information, but also provides a more efficient and intelligent solution for data processing and analysis of meteorological service platforms. Through the technical solutions of the embodiments of this application, it is possible to achieve the precise construction of structured data for the interaction information of the meteorological service platform, providing strong support for subsequent data analysis, prediction, and services.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method and system for data management of a meteorological service platform based on artificial intelligence. Background Art

[0002] In the field of meteorological services, with the continuous increase of user interaction information, how to efficiently and accurately process this information, extract key meteorological content elements, and construct structured data has always been a technical problem to be solved urgently. Traditional data processing methods are often inefficient when dealing with complex and unstructured user interaction information, and it is difficult to accurately extract key meteorological elements and their relationships. Summary of the Invention

[0003] To improve the above problems, this application provides a method and system for data management of a meteorological service platform based on artificial intelligence.

[0004] An embodiment of this application provides a method for data management of a meteorological service platform based on artificial intelligence, which is applied to a big data management system. The method for data management of the meteorological service platform based on artificial intelligence includes:

[0005] Obtain a network learning example set of a structured data generation network to be debugged, and perform dense interaction semantic mining on the meteorological service platform interaction information samples in the network learning example set to obtain dense interaction semantic samples of the meteorological service platform interaction information samples;

[0006] Extract meteorological content element samples from the dense interaction semantic samples to obtain each meteorological content element sample in the meteorological service platform interaction information samples, and determine the relationship state evaluation weight of the meteorological content element samples based on the dense vectors of the meteorological elements of the meteorological content element samples;

[0007] Determine the meteorological content element discrimination training error of the meteorological content element samples based on the dense vectors of the meteorological elements of the meteorological content element samples and the relationship state description features of the meteorological service platform interaction information samples;

[0008] Extract target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element samples, and determine the target discrimination training error of the meteorological service platform interaction information samples based on the dense vectors of the meteorological elements of the target meteorological content element samples and the dense interaction semantic samples;

[0009] Debug the to-be-debugged structured data generation network based on the target discrimination training error to obtain a target structured data generation network, so as to refine the structured relationship status of the interaction information of the meteorological service platform through the target structured data generation network; wherein, the refined structured relationship status is used to implement the structured data construction of the interaction information of the meteorological service platform.

[0010] Optionally, extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship status evaluation weights from the meteorological content element samples includes:

[0011] Extracting binary groups of meteorological content element samples from the meteorological content element samples;

[0012] Based on the meteorological content element discrimination training errors and relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples, extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship status evaluation weights from the meteorological content element samples.

[0013] Optionally, the extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship status evaluation weights from the meteorological content element samples based on the meteorological content element discrimination training errors and relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples includes:

[0014] Based on the meteorological content element discrimination training errors of the meteorological content element samples in the binary groups of meteorological content element samples, determining a first comparison result between the relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples;

[0015] Based on the relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples, determining a second comparison result between the relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples;

[0016] If the first comparison result does not match the second comparison result, determining the meteorological content element samples in the binary groups of meteorological content element samples as target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship status evaluation weights.

[0017] Optionally, the extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship status evaluation weights from the meteorological content element samples based on the meteorological content element discrimination training errors and relationship status evaluation weights of the meteorological content element samples in the binary groups of meteorological content element samples includes:

[0018] Determine the relationship status sorting error of the meteorological content element sample in the meteorological content element sample based on the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element sample in the meteorological content element sample binary group;

[0019] Extract the meteorological content element sample corresponding to the relationship status sorting error that meets the set requirements from the meteorological content element sample to obtain the target meteorological content element sample with unmatched discrimination training error and relationship status evaluation weight.

[0020] Optionally, the determining the relationship status sorting error of the meteorological content element sample in the meteorological content element sample based on the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element sample in the meteorological content element sample binary group includes:

[0021] Determine the discrimination error comparison result between the meteorological content element samples in the meteorological content element sample binary group based on the meteorological content element discrimination training error of the meteorological content element sample in the meteorological content element sample binary group;

[0022] Determine the evaluation weight comparison result between the meteorological content element samples in the meteorological content element sample binary group based on the relationship status evaluation weight of the meteorological content element sample in the meteorological content element sample binary group;

[0023] Determine the relationship status sorting error of the meteorological content element sample in the meteorological content element sample based on the discrimination error comparison result and the evaluation weight comparison result.

[0024] Optionally, the determining the relationship status sorting error of the meteorological content element sample in the meteorological content element sample based on the discrimination error comparison result and the evaluation weight comparison result includes:

[0025] Perform weighting on the discrimination error comparison result and the evaluation weight comparison result to obtain a weighted comparison result;

[0026] Obtain the target loss parameter, and determine the relationship status sorting error of the meteorological content element sample in the meteorological content element sample based on the target loss parameter and the weighted comparison result.

[0027] Optionally, the extracting the meteorological content element sample corresponding to the relationship status sorting error that meets the set requirements from the meteorological content element sample to obtain the target meteorological content element sample with unmatched discrimination training error and relationship status evaluation weight includes:

[0028] If the relationship status priority of the meteorological content element sample is paired with the weighted comparison result, it is determined that the relationship status sorting error of the meteorological content element sample meets the set requirements;

[0029] Use the meteorological content element sample corresponding to the relationship status sorting error that meets the set requirements as the target meteorological content element sample.

[0030] Optionally, the extracting of meteorological content elements from the dense interaction semantic sample to obtain each meteorological content element sample in the meteorological service platform interaction information sample includes:

[0031] Extract each initial meteorological content element sample in the meteorological service platform interaction information sample by performing meteorological content element extraction on the dense interaction semantic sample;

[0032] Extract meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples.

[0033] Optionally, the extracting of meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples includes:

[0034] Determine the initial relationship status evaluation weight of the initial meteorological content element sample based on the initial meteorological element dense vector of the initial meteorological content element sample;

[0035] Extract the initial meteorological content element sample with the largest initial relationship status evaluation weight from the initial meteorological content element samples;

[0036] Determine the overlap coefficient between the content element coverage vector of the initial meteorological content element sample and the content element coverage vector of the initial meteorological content element sample with the largest initial relationship status evaluation weight;

[0037] If the overlap coefficient is less than the set coefficient, use the initial meteorological content element sample corresponding to the overlap coefficient as the meteorological content element sample.

[0038] Optionally, the performing of dense interaction semantic mining on the meteorological service platform interaction information samples in the network learning example set to obtain the dense interaction semantic samples of the meteorological service platform interaction information samples includes: generating a first dense interaction semantic mining branch in the to-be-debugged structured data generation network, and performing dense interaction semantic mining on the meteorological service platform interaction information samples in the network learning example set to obtain the dense interaction semantic samples of the meteorological service platform interaction information samples;

[0039] Before determining the meteorological content element discrimination training error of the meteorological content element example based on the relationship state description feature between the dense vector of the meteorological element of the meteorological content element example and the interaction information example of the meteorological service platform, it further includes: generating a second dense interaction semantic mining branch in the to-be-debugged structured data generation network, performing dense interaction semantic mining on the meteorological content element example, and obtaining the dense vector of the meteorological element of the meteorological content element example.

[0040] Optionally, after debugging the to-be-debugged structured data generation network based on the target discrimination training error to obtain a target structured data generation network, the method further includes:

[0041] Performing dense interaction semantic mining on the interaction information of the to-be-managed meteorological service platform through the target structured data generation network to obtain the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform;

[0042] Based on the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform, determining the local meteorological content elements in the interaction information of the to-be-managed meteorological service platform;

[0043] Based on the dense vector of the meteorological element of the local meteorological content element and the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform, refining the structured relationship state of the interaction information of the to-be-managed meteorological service platform to obtain the current structured relationship state of the interaction information of the to-be-managed meteorological service platform.

[0044] An embodiment of the present application provides a big data management system, including at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the above method.

[0045] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program implements the above method when running.

[0046] Through in-depth mining of the semantic content of the interaction information in the embodiments of the present application, the embodiments of the present application can more accurately refine meteorological elements, and determine the relationship state evaluation weight based on the dense vector of the meteorological element, so as to obtain a more accurate meteorological content element discrimination training error. This innovative method not only solves the deficiencies of traditional methods in processing unstructured information, but also provides a more efficient and intelligent solution for data processing and analysis of meteorological service platforms. Through the technical solution of the embodiments of the present application, precise structured data construction of the interaction information of the meteorological service platform can be realized, providing strong support for subsequent data analysis, prediction and services. Description of the Drawings

[0047] Figure 1 The flowchart of a data management method for a meteorological service platform based on artificial intelligence provided by an embodiment of the present application.

[0048] Figure 2 The schematic structural diagram of a big data management system 200 provided by an embodiment of the present application. Specific implementation manners

[0049] In order to better understand the above technical solution, the technical solution of the present application will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solution of the present application, rather than limitations on the technical solution of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0050] Figure 1 A data management method for a meteorological service platform based on artificial intelligence is shown, which is applied to a big data management system. The method includes the following steps 101-step 109.

[0051] Step 101, obtain a network learning example set for a structured data generation network to be debugged, and perform dense interaction semantic mining on the meteorological service platform interaction information samples in the network learning example set to obtain dense interaction semantic samples of the meteorological service platform interaction information samples.

[0052] Among them, the structured data generation network to be debugged is a neural network model that has not been optimized or debugged, and is specifically designed to process and generate structured data. In the scenario of meteorological services, the purpose of this network is to extract useful structured data from unstructured meteorological service platform interaction information, such as specific meteorological parameters such as temperature, humidity, and wind speed. After being processed by the network, these data can be used more effectively for meteorological analysis, prediction, and services.

[0053] The network learning example set is a data set containing multiple samples, which is used to train and optimize the structured data generation network to be debugged. In the context of meteorological services, the network learning example set may include various historical meteorological data, interaction records between users and meteorological service platforms, etc. These data are used as "textbooks" to help the network learn how to extract key structured data from complex interaction information.

[0054] The meteorological service platform interaction information samples can be understood as data samples generated when users perform queries, feedback, or other interaction behaviors on the meteorological service platform. These data not only include the text, voice, or image information input by users, but may also include users' click behaviors, browsing histories, etc. In the big data management system, these interaction information samples are important data sources for understanding user needs and optimizing meteorological services.

[0055] Dense interaction semantic mining is used to mine deep semantic information from sample interaction information of a meteorological service platform. Through advanced algorithms such as natural language processing and machine learning, the big data management system can identify and understand the implicit meaning and context relationship in user interactions, so as to generate a richer data representation. This kind of mining helps the system to understand user needs more accurately and provide personalized meteorological services.

[0056] Dense interaction semantic samples are data samples obtained after dense interaction semantic mining. These samples not only retain the core content of the original interaction information, but also enhance the semantic information in it, enabling the big data management system to more easily identify key meteorological content elements. Dense interaction semantic samples are an important basis for subsequent data processing and model training.

[0057] Furthermore, dense interaction semantic samples (dense interaction semantic feature samples) play an important role in modern meteorological services. The following is a detailed introduction to this concept through specific feature value examples.

[0058] (1) Temperature feature: In a certain user interaction, the user queried "What's the weather like in Beijing tomorrow?". Through dense interaction semantic mining, the system can identify keywords such as "tomorrow", "Beijing", and "weather", and further analyze the user's intention. For the temperature feature, the system may generate a specific value, such as "22 degrees", to represent the highest temperature in Beijing tomorrow. This value is a feature value in the dense interaction semantic sample.

[0059] (2) Humidity feature: In the above query, in addition to providing temperature information, the system can also give humidity information. Through dense interaction semantic mining, the system can analyze the humidity range in Beijing tomorrow, such as "40%-60%". The numerical value of this humidity range is also an important feature in the dense interaction semantic sample.

[0060] (3) Wind direction and wind speed feature: When the user queries the weather, wind direction and wind speed are also key concerns of the user. Through dense interaction semantic mining, the system can give the specific wind direction (such as "southeast wind") and wind speed (such as "level 3-4"). These wind direction and wind speed information will also be reflected in the dense interaction semantic sample in numerical form (such as the numerical value corresponding to the wind direction angle and wind speed level).

[0061] (4) Weather condition characteristics: Weather conditions (such as sunny, rainy, snowy, etc.) are one of the most concerned information when users query the weather. Through dense interactive semantic mining, the system can identify weather conditions and convert them into corresponding numerical representations. For example, "sunny" can be represented as "0", "rainy" as "1", "snowy" as "2", etc. These numerical values constitute the weather condition characteristics in the dense interactive semantic examples.

[0062] It can be seen that the dense interactive semantic examples are obtained by deeply mining and analyzing the interaction information between users and the meteorological service platform. These examples represent each feature in the interaction information in the form of precise numerical values, such as temperature, humidity, wind direction and speed, and weather conditions. These numerical features provide strong support for subsequent data processing, model training, and meteorological services.

[0063] Specifically, step 101 involves two main operations: obtaining a network learning example set and performing dense interactive semantic mining on the meteorological service platform interaction information examples. First, the big data management system obtains the network learning example set required for the network from multiple sources for debugging structured data. These data may come from historical meteorological observation records, real-time interaction data between users and the meteorological service platform, and other relevant data sources. These data sets contain rich meteorological information and user behavior data, which are the key to training and optimizing the data generation network. Next, the system performs dense interactive semantic mining on the meteorological service platform interaction information examples in these network learning example sets. This process aims to deeply understand and extract the deep meanings and context relationships in these interaction information. By using advanced technologies such as natural language processing and deep learning, the system can identify keywords, phrases, and intentions in user queries, as well as the complex relationships between them. During the dense interactive semantic mining process, the big data management system generates dense interactive semantic examples. These examples not only contain the core content of the original interaction information, but also enhance the semantic information in them through methods such as vector representation. This enables the system to more easily identify key meteorological content elements and provides an important basis for subsequent data processing and model training. Therefore, step 101 is a key step in the big data management system's processing of meteorological service platform interaction information. It lays a solid foundation for the training and debugging of the subsequent structured data generation network by obtaining a rich network learning example set and performing in-depth semantic mining.

[0064] Step 103: Extract meteorological content elements from the dense interactive semantic examples to obtain each meteorological content element example in the meteorological service platform interaction information example, and determine the relationship status evaluation weight of the meteorological content element example based on the meteorological element dense vector of the meteorological content element example.

[0065] Among them, the extraction of meteorological content elements can be understood as identifying and separating key meteorological information elements from a large amount of meteorological service platform interaction information. These elements include, but are not limited to, temperature, humidity, wind direction, wind speed, weather conditions (sunny, rainy, snowy, etc.), visibility, etc. The purpose of extracting these elements is to provide more precise customized meteorological services for users, and at the same time provide a basis for the analysis and prediction of meteorological data.

[0066] A sample of meteorological content elements can be understood as a specific instance obtained through the extraction of meteorological content elements. Represented by a numerical feature vector. For example, a sample of meteorological content elements can be a vector containing multiple feature values, such as [temperature: 25°C, humidity: 60%, wind direction: southeast, wind speed: 3m / s, weather condition: sunny, visibility: 10km]. This sample reflects the specific meteorological conditions at a certain place and time, and each feature value is accurately represented in numerical form.

[0067] A dense vector of meteorological elements is a compact numerical representation of a sample of meteorological content elements. It integrates multiple meteorological elements into a high-dimensional vector, and each dimension represents a specific meteorological feature. For example, a dense vector of meteorological elements may be as follows: [0.8, -0.2, 0.5, 0.3, -0.1,...], where each numerical value represents the value of a specific meteorological feature (such as temperature, humidity, etc.) after a certain transformation (such as normalization, standardization, etc.). This vector representation helps with subsequent data processing and model training.

[0068] The evaluation weight of the relationship state can be understood as the relative importance occupied by each element when evaluating the relationship or state between samples of meteorological content elements. These weights can be determined based on historical data, expert knowledge, or machine learning algorithms. For example, when predicting future weather conditions, temperature and humidity may be more important than wind direction and wind speed, so they will be given higher weights.

[0069] Specifically, the purpose of step 103 is to extract meteorological content elements from the dense interaction semantic examples and determine the weights of these elements in subsequent analysis or prediction. First, the system uses natural language processing and machine learning technologies to deeply analyze the dense interaction semantic examples. This process involves identifying key meteorological words, phrases, and context information in the text to accurately extract key meteorological content elements such as temperature, humidity, wind direction and speed, and weather conditions. The extracted meteorological content elements will be converted into numerical feature vectors, that is, meteorological content element examples. These examples represent the states and changes of various meteorological features in an accurate numerical form. For example, the system may extract a vector containing multiple feature values such as temperature, humidity, and wind direction to represent the meteorological conditions at a certain moment. Next, the system generates dense vectors of meteorological elements based on these meteorological content element examples. This process involves integrating multiple meteorological features into a high-dimensional vector for subsequent data processing and model training. This representation of dense vectors helps to capture the complex relationships and dependencies between meteorological features. Finally, the system determines the relationship state evaluation weights for each meteorological content element example. These weights reflect the importance and influence of each element when evaluating meteorological conditions or making predictions. The determination of weights may be based on historical data, expert judgment, or the output of machine learning algorithms. For example, when predicting future weather changes, the system may assign higher weights to temperature and humidity because they play a more critical role in weather changes. Through the above steps, the system can accurately extract key meteorological content elements from the dense interaction semantic examples and provide strong support for subsequent data analysis, model training, and meteorological services.

[0070] It can be seen that one of the ideas of the embodiments of this application is to construct an efficient and intelligent big data management system, which can use the target structured data generation network optimized and debugged to deeply refine the structured relationship state of the user interaction information on the meteorological service platform. This refining process aims to identify and extract key meteorological elements and their mutual relationships from unstructured text data, and further convert these complex information into a clear and orderly structured data format. In this way, the structured management of the interaction information on the meteorological service platform can be realized, laying a solid foundation for subsequent data processing and analysis work.

[0071] Specifically, the embodiments of the present application innovatively use a target structured data generation network, which is a deep learning model that can self-learn through a large amount of training data and continuously optimize its ability to extract structured information from complex texts. The application of this network is novel in the field of meteorological data management, greatly improving the level of data processing intelligence. In addition, traditional data processing methods often have difficulty effectively processing unstructured text data, while the embodiments of the present application can convert unstructured user interaction information into structured data through structured relationship state refinement, which not only simplifies the data processing process but also improves the usability and analyzability of the data. Then, by implementing the structured management of interaction information on the meteorological service platform, the embodiments of the present application greatly improve the efficiency of data automatic classification, storage, analysis, and prediction. In addition, the use of structured data also significantly improves the speed and accuracy of data retrieval, thus enabling the provision of more convenient and efficient meteorological data services for users. In the field of meteorological services, rapid and accurate data analysis and prediction are crucial for timely response and decision-making. The embodiments of the present application enable more scientific and accurate data-based decision-making by providing high-quality structured data.

[0072] In summary, the embodiments of the present application demonstrate remarkable innovation in the field of meteorological data management. By introducing an advanced structured data generation network, they achieve the efficient processing and analysis of unstructured text data, providing strong data support for meteorological services and promoting the intelligent and efficient development of meteorological services.

[0073] Step 105: Determine the meteorological content element discrimination training error of the meteorological content element example based on the meteorological element dense vector of the meteorological content element example and the relationship state description features of the meteorological service platform interaction information example.

[0074] Among them, the relationship state description features refer to the features that can depict the relationships between meteorological content elements or between meteorological content and user interaction intentions. These features can be represented by numerical feature vectors to capture the correlation, dependence, and change trends between different elements. For example, a relationship state description feature vector may contain multiple dimensions, and each dimension represents the relationship state between different meteorological elements, such as "the correlation between temperature and humidity" and "the dependence of wind direction on weather conditions". Such a feature vector may be as follows: [0.7, -0.3, 0.5,...], where each value represents the intensity or degree of a specific relationship state.

[0075] The training error of meteorological content element discrimination refers to the difference between the model prediction result and the actual label when training the meteorological content element discrimination model. This error is usually measured by a loss function, which can quantify the accuracy of the model prediction. A commonly used loss function is the cross-entropy loss, which measures the difference between the probability distribution predicted by the model and the actual label. For example, in a multi-classification problem, if the model predicts the probability of a certain meteorological element belonging to a certain class as 0.8, while the actual label shows that the element does not belong to this class, then the cross-entropy loss will be relatively large, indicating a large error in this prediction of the model.

[0076] Specifically, step 105 is a key step in evaluating and optimizing the meteorological content element discrimination model. The purpose of this step is to determine the training error of the model in discriminating meteorological content elements based on the meteorological element dense vectors of meteorological content element examples and the relationship state description features of meteorological service platform interaction information examples. First, the system obtains the meteorological element dense vectors and relationship state description features of the meteorological content element examples that have been extracted. These feature vectors and description features are the basic data for the model to make discrimination. Then, the system inputs these feature vectors and description features into the meteorological content element discrimination model that has been constructed. This model may be a classifier based on deep learning or other machine learning techniques, which is used to predict the classification or state of meteorological content elements according to the input feature vectors. The model will output the prediction results of each meteorological content element example, and these results are usually represented in the form of a probability distribution. For example, for a certain meteorological element (such as weather condition), the model may give the probabilities of belonging to different categories (clear, rainy, snowy, etc.). Then, the system compares these prediction results with the actual labels to calculate the training error. The error measurement criterion used here is usually a loss function, such as cross-entropy loss, etc. The loss function can quantify the accuracy of the model prediction. The smaller the loss value, the more accurate the model prediction. Finally, the system records and analyzes these training errors for subsequent optimization of the model. By analyzing the source and distribution of the errors, the system can adjust the parameters and structure of the model to improve the accuracy of the model in discriminating meteorological content elements. Through the above steps, the system can evaluate and optimize the performance of the meteorological content element discrimination model, providing more accurate and reliable data support for subsequent meteorological services.

[0077] Step 107: Extract the target meteorological content element examples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element examples, and determine the target discrimination training error of the meteorological service platform interaction information examples based on the meteorological element dense vectors of the target meteorological content element examples and the dense interaction semantic examples.

[0078] Among them, the target meteorological content element samples can be understood as those meteorological content element samples that are specifically selected for further analysis or model training in the meteorological service platform interaction information. These samples are usually representative and can reflect a certain specific meteorological phenomenon, user interaction intention or business requirement. For example, when predicting extreme weather events (such as heavy rain, snowstorm), the meteorological content element samples closely related to such events may be selected as target samples. These samples not only contain key meteorological features (such as sudden temperature drop, humidity increase, wind direction change, etc.), but may also involve users' inquiries or feedback on these feature changes, thus providing richer context information for the model.

[0079] The target discrimination training error can be understood as the difference between the model prediction result and the actual result when using the target meteorological content element samples for model training. This error is quantified by a specific loss function and is used to evaluate the performance of the model on a specific task. Taking the classification task as an example, the commonly used loss function includes the cross-entropy loss (Cross-Entropy Loss), which measures the difference between the probability distribution predicted by the model and the true label. The specific calculation formula is: (L=-sum_{i}y_ilog(p_i)), where (y_i) is the true label (usually 0 or 1), and (p_i) is the probability predicted by the model. When the model prediction is completely accurate, the cross-entropy loss is 0; the more inaccurate the prediction, the larger the loss value. The target discrimination training error is calculated based on such a loss function and is used to guide the optimization direction of the model.

[0080] In step 107, focus on identifying and extracting from existing meteorological content element samples those target meteorological content element samples where the discriminative training error of meteorological content elements and the relationship state evaluation weight do not match. These samples are crucial for subsequent model optimization and improving prediction accuracy. First, the system reviews and analyzes the discriminative training error of meteorological content elements and the relationship state evaluation weight calculated in step 105. By comparing the matching degree of the two, the system can identify samples with relatively large errors but small weights, or samples with relatively small errors but large weights, which are the so-called unmatched samples. Next, the system further screens out representative target meteorological content element samples from these unmatched samples. These samples may contain special meteorological phenomena that are difficult for the model to accurately predict, rare intentions in user interactions, or key points in business requirements. Once the target meteorological content element samples are selected, the system recalculates the discriminative training error based on the dense vectors of meteorological elements and the dense interaction semantic samples of these samples. This process involves inputting the target samples into the trained model, obtaining the prediction results of the model, and comparing them with the actual labels to calculate the error. At this time, the system uses a specific loss function (such as the cross-entropy loss function) to quantify this error, thereby obtaining the target discriminative training error. This error value reflects the performance of the model when processing these specific samples, providing an important basis for subsequent model optimization. Finally, the system records and analyzes these target discriminative training errors to identify areas where the model's prediction ability needs to be improved. Based on these analysis results, the system can adjust the model's parameters, structure, or training strategy to improve the accuracy and robustness of the model in future predictions.

[0081] Step 109: Debug the to-be-debugged structured data generation network based on the target discriminative training error to obtain a target structured data generation network, so as to refine the structured relationship state of the meteorological service platform interaction information through the target structured data generation network; wherein, the refined structured relationship state is used to implement the structured data construction of the meteorological service platform interaction information.

[0082] Among them, the target structured data generation network can be understood as a deep learning network model optimized and debugged in modern meteorological services, which can effectively extract structured data from complex meteorological service platform interaction information. This network model is usually based on natural language processing and machine learning technologies, and learns how to identify and parse key information in text through a large amount of training data, and then converts this information into a structured data format. The target structured data generation network not only improves the efficiency and accuracy of data processing, but also provides convenience for subsequent data analysis and meteorological services.

[0083] Structured relationship status refinement can be understood as identifying and extracting key meteorological elements and their relationship status from the interaction information of the meteorological service platform, and representing this information in a structured form. This process usually relies on advanced natural language processing and machine learning algorithms, which can automatically identify entities, attributes, and relationships in the text, and then construct a clear, easy-to-understand, and analyzable structured data model. Through structured relationship status refinement, the connections and impacts between different meteorological elements can be understood more intuitively, providing strong data support for meteorological prediction, analysis, and services.

[0084] The construction of structured data for the interaction information of the meteorological service platform can be understood as the process of converting the interaction information of users on the meteorological service platform, such as queries, feedbacks, comments, etc., into structured data. This kind of structured data is not only convenient for storage and management, but also facilitates subsequent data mining and analysis. By constructing structured data, the needs and preferences of users can be understood more clearly, optimizing the quality and efficiency of meteorological services. At the same time, structured data also makes it possible to develop more intelligent meteorological service applications, such as personalized recommendations based on user behavior, visual display of meteorological data, etc.

[0085] Therefore, in step 109, the big data management system conducts meticulous debugging and optimization on the structured data generation network to be debugged based on the target discrimination training error calculated in the previous steps. First, the big data management system receives and analyzes the target discrimination training error. These error data provide crucial feedback on the performance of the network model, indicating which aspects of the model's prediction and parsing capabilities need to be improved. Next, the management system uses these error data to adjust the parameters, optimize the structure, or improve the learning strategy of the structured data generation network to be debugged. For example, it may adjust the weights and biases in the network, or increase the number of network layers to enhance the complexity of the model and make it better fit the data. After a series of debugging and optimization, the management system obtains a target structured data generation network with better performance. This network can more accurately identify and extract key meteorological elements and their relationship states when parsing the interaction information of the meteorological service platform. Subsequently, the management system uses this target structured data generation network to refine the structured relationship state of the interaction information on the meteorological service platform. In this process, the network model automatically analyzes the text data, identifies various meteorological elements such as temperature, humidity, wind direction, etc., and clarifies their relationship states, such as "an increase in temperature leads to a decrease in humidity", etc. Finally, the refined structured relationship state is used to implement the construction of structured data for the interaction information of the meteorological service platform. This means that the originally scattered and unstructured text data is converted into a clear and orderly structured data format. This format not only facilitates data storage and management but also provides great convenience for subsequent data analysis, meteorological prediction, and services. It can be seen that step 109 is a key link in data processing and analysis in modern meteorological services. By debugging and optimizing the structured data generation network, it achieves the goal of refining the structured relationship state from complex text data, laying a solid foundation for the intelligence and efficiency of meteorological services.

[0086] Combined with the above content, in modern meteorological services, the big data management system plays a crucial role. The following will introduce in detail how the big data management system deeply processes the interaction information of the meteorological service platform and realizes the construction of structured data through a complete and specific application scenario.

[0087] First, in step 101, the big data management system obtains the structured data corresponding to the structured data generation network to be debugged from various channels, and these structured data constitute a network learning example set. The network learning example set can come from various sources such as meteorological observation stations, satellite remote sensing, user feedback, etc. The big data management system conducts in-depth semantic mining on the interaction information of the meteorological service platform in these data, and this process is called dense interaction semantic mining. Through this mining, the big data management system can understand the deep meaning in the information, thereby generating dense interaction semantic examples.

[0088] Next, in step 103, the big data management system further processes these dense interaction semantic examples. The big data management system extracts elements related to meteorological content, such as temperature, humidity, wind speed, etc., which are called meteorological content element examples. The big data management system also assigns a relationship status evaluation weight to each element example based on the dense vectors of meteorological elements of these elements. The relationship status evaluation weight reflects the importance of the element in the overall meteorological information.

[0089] Then, entering step 105, the big data management system uses advanced algorithms, combines the dense vectors of meteorological elements of meteorological content element examples and the relationship status description features of the original meteorological service platform interaction information examples to determine the discriminant training error of each meteorological content element example. This error reflects the accuracy of the big data management system in identifying and interpreting these elements.

[0090] Furthermore, in step 107, the big data management system carefully analyzes the data obtained in the previous steps. It will find out those target meteorological content element examples where the discriminant training error of meteorological content elements and the relationship status evaluation weight do not match. These examples may be parts that the big data management system has misunderstood or omitted during the processing. Then, the big data management system determines the target discriminant training error of the overall meteorological service platform interaction information example based on the dense vectors of meteorological elements and the dense interaction semantic examples of these target examples.

[0091] Finally, in step 109, the big data management system uses the above-mentioned target discriminant training error to finely debug the structured data generation network to be debugged. This process is to improve the accuracy and efficiency of the big data management system in processing similar information in the future. After the debugging is completed, the big data management system will obtain a target structured data generation network, which can efficiently refine the structured relationship status of the meteorological service platform interaction information. The refined structured relationship status is crucial for realizing the structured data construction of the meteorological service platform interaction information, and it will help to improve the overall quality and user experience of meteorological services.

[0092] In summary, the big data management system plays a core role in the processing of meteorological service platform interaction information. Through a series of complex processes such as in-depth mining, element extraction, error determination, and network debugging, it finally achieves the accurate grasp and efficient utilization of meteorological information.

[0093] It can be seen that in the embodiments of the present application, by deeply mining the dense interaction semantics of the meteorological service platform interaction information, meteorological content elements are accurately extracted, and meteorological element dense vectors are innovatively used to determine the relationship state evaluation weights, so as to obtain a more accurate discriminant training error of meteorological content elements. Further, by screening out the target meteorological content element samples where the discriminant training error does not match the relationship state evaluation weight, the embodiments of the present application can more carefully determine the target discriminant training error, and then accurately debug the structured data generation network to be debugged, and finally obtain an efficient and accurate target structured data generation network. This innovative method not only improves the accuracy of refining the structured relationship state of the meteorological service platform interaction information, but also provides strong technical support for the construction of structured data of the meteorological service platform interaction information, greatly improving the intelligent level and efficiency of data processing.

[0094] In some alternative embodiments, extracting the target meteorological content element samples where the discriminant training error of the meteorological content elements and the relationship state evaluation weight do not match from the meteorological content element samples includes: extracting the meteorological content element sample binary groups from the meteorological content element samples; based on the discriminant training error of the meteorological content elements and the relationship state evaluation weight of the meteorological content element samples in the meteorological content element sample binary groups, extracting the target meteorological content element samples where the discriminant training error of the meteorological content elements and the relationship state evaluation weight do not match from the meteorological content element samples.

[0095] In the application scenario of digital meteorological services, the big data management system will extract the target meteorological content element samples where the discriminant training error of the meteorological content elements and the relationship state evaluation weight do not match from the meteorological content element samples.

[0096] First, the big data management system extracts the meteorological content element sample binary groups from the previously processed meteorological content element samples. Here, the "binary group" refers to a combination composed of two elements, that is, each meteorological content element sample will be paired with its corresponding discriminant training error of the meteorological content elements and the relationship state evaluation weight.

[0097] Next, the system conducts a detailed analysis of these binary groups. It will compare the discriminant training error of the meteorological content elements and the relationship state evaluation weight of the meteorological content element samples in each binary group to determine whether there is a mismatch between the two. If there is a mismatch, that is, the difference between the discriminant training error and the evaluation weight exceeds the preset threshold, then such a meteorological content element sample will be identified as a "target meteorological content element sample" by the system.

[0098] For example, if the discrimination training error of a sample of meteorological content elements is very high, but its relationship status evaluation weight is very low, or vice versa, this indicates a large deviation between the performance of this sample during the training process and its actual importance. Such samples have important reference value for optimizing the structured data generation network of the big data management system.

[0099] Once these target meteorological content element samples are identified, the big data management system will perform special processing on them. The system may adjust the weights of these samples in subsequent training, or conduct a more in-depth analysis of them to find out the root causes of the mismatch, and optimize the performance and accuracy of the structured data generation network accordingly.

[0100] Through such a processing flow, the big data management system can continuously self-optimize and improve, so as to better serve the needs of digital meteorological services.

[0101] It can be seen that in the application of digital meteorological services, the big data management system can specifically optimize the training process of its structured data generation network by accurately extracting the target meteorological content element samples with mismatched discrimination training errors and relationship status evaluation weights of meteorological content elements. This optimization not only improves the network's processing ability for complex meteorological data, but also enables the system to more accurately extract key meteorological elements and their mutual relationships from user interaction information. Therefore, this embodiment significantly improves the intelligent level and data processing efficiency of the big data management system in meteorological services, providing users with more accurate and efficient meteorological data services.

[0102] In the following steps, extracting the target meteorological content element samples with mismatched discrimination training errors and relationship status evaluation weights from the meteorological content element samples based on the discrimination training errors and relationship status evaluation weights of the meteorological content element samples in the meteorological content element sample pair includes: determining a first comparison result between the relationship status evaluation weights of the meteorological content element samples in the meteorological content element sample pair based on the discrimination training errors of the meteorological content element samples in the meteorological content element sample pair; determining a second comparison result between the relationship status evaluation weights of the meteorological content element samples in the meteorological content element sample pair based on the relationship status evaluation weights of the meteorological content element samples in the meteorological content element sample pair; if the first comparison result does not match the second comparison result, then determining the meteorological content element samples in the meteorological content element sample pair as the target meteorological content element samples with mismatched discrimination training errors and relationship status evaluation weights.

[0103] In the application scenario of digital meteorological services, a big data management system needs to process and analyze a large amount of meteorological data to provide accurate meteorological forecasts and services. In this embodiment, the system will extract the target meteorological content element samples where the two do not match according to the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element samples.

[0104] First, the big data management system extracts the meteorological content element sample pairs from the processed meteorological content element samples. These pairs contain the meteorological content element samples and their corresponding meteorological content element discrimination training errors and relationship status evaluation weights.

[0105] Next, the system performs two comparison operations. The first comparison is based on the meteorological content element discrimination training error. The system calculates the discrimination training error of each meteorological content element sample and determines the first comparison result of the relationship status evaluation weights between them according to this error. Simply put, this step is to judge which samples have relatively higher or lower discrimination training errors. The second comparison is based on the relationship status evaluation weight. The system checks the relationship status evaluation weight of each meteorological content element sample and determines the second comparison result between them. This step is to understand which samples are more important or less important in the relationship status.

[0106] Finally, the system compares the results of these two comparisons. If a meteorological content element sample shows a relatively high discrimination training error in the first comparison but is given a relatively low relationship status evaluation weight in the second comparison, or vice versa, then this sample is considered "unpaired". In other words, the discrimination training error of this sample does not match its importance in the relationship status. Once these unpaired target meteorological content element samples are identified, the big data management system can further analyze and process them, such as adjusting the training strategy, optimizing the model parameters, etc., to improve the accuracy and efficiency of meteorological data processing.

[0107] Through this embodiment, the big data management system can more accurately identify the meteorological content element samples where there is a mismatch between the discrimination training error and the relationship status evaluation weight. The identification of this mismatch helps the system discover potential problems in the data processing and analysis process, and then optimize the model and algorithm to improve the accuracy and reliability of meteorological forecasts and services. At the same time, this refined data processing method also helps to improve the overall performance and intelligent level of the big data management system in digital meteorological services.

[0108] In some preferred embodiments, based on the meteorological content element discrimination training error and the relationship state evaluation weight of the meteorological content element sample in the meteorological content element sample binary group, extracting the target meteorological content element sample with unmatched meteorological content element discrimination training error and relationship state evaluation weight from the meteorological content element sample includes: based on the meteorological content element discrimination training error and the relationship state evaluation weight of the meteorological content element sample in the meteorological content element sample binary group, determining the relationship state sorting error of the meteorological content element sample in the meteorological content element sample; extracting the meteorological content element sample corresponding to the relationship state sorting error that meets the set requirements from the meteorological content element sample, so as to obtain the target meteorological content element sample with unmatched meteorological content element discrimination training error and relationship state evaluation weight.

[0109] In the application scenario of digital meteorological services, the data processed by the big data management system is complex and diverse, including various meteorological content element samples. To improve the accuracy and efficiency of data processing, the system needs to perform refined analysis and screening on these samples.

[0110] In this preferred embodiment, the big data management system first focuses on the meteorological content element sample binary group, which consists of the meteorological content element sample, its corresponding meteorological content element discrimination training error, and the relationship state evaluation weight. The system conducts in-depth analysis on these binary groups to determine the relationship state sorting error of each sample.

[0111] The relationship state sorting error is a comprehensive index that considers the matching degree between the discrimination training error of the sample and the relationship state evaluation weight. For example, if a certain sample has a high discrimination training error but a relatively low relationship state evaluation weight, or vice versa, then the relationship state sorting error of this sample will be relatively large. This error reflects the inconsistency between the performance of the sample in the training process and its importance in the relationship state.

[0112] To calculate the relationship state sorting error more precisely, the big data management system may adopt a weighted error calculation method. For example, the system can perform a certain form of weighted summation on the discrimination training error and the relationship state evaluation weight to obtain a comprehensive error index. This weighted summation process can be represented by a formula, such as: relationship state sorting error = α * discrimination training error + β * (1 - relationship state evaluation weight), where α and β are weighted coefficients and can be adjusted according to the actual situation.

[0113] After calculating the relationship status sorting error, the big data management system extracts meteorological content element samples that meet the set requirements based on this error. For example, the system may set an error threshold. When the relationship status sorting error of a certain sample exceeds this threshold, it is considered that the discrimination training error and the relationship status evaluation weight of this sample do not match, and thus it is extracted as the target meteorological content element sample.

[0114] Through this preferred embodiment, the big data management system can more accurately identify and extract those meteorological content element samples whose discrimination training errors and relationship status evaluation weights do not match. This not only helps the system discover potential problems in the data processing process, but also provides a targeted improvement direction for subsequent model training and algorithm optimization. At the same time, through the method of weighted calculation of the relationship status sorting error, the system can more comprehensively evaluate the comprehensive performance of the samples, thereby improving the accuracy and reliability of digital meteorological services. Overall, this embodiment significantly improves the intelligent level and efficiency of the big data management system in meteorological data processing and analysis.

[0115] In the next step, determining the relationship status sorting error of the meteorological content element sample in the meteorological content element sample pair based on the discrimination training error of the meteorological content element sample and the relationship status evaluation weight in the meteorological content element sample pair includes: determining the discrimination error comparison result between the meteorological content element samples in the meteorological content element sample pair based on the discrimination training error of the meteorological content element sample in the meteorological content element sample pair; determining the evaluation weight comparison result between the meteorological content element samples in the meteorological content element sample pair based on the relationship status evaluation weight of the meteorological content element sample in the meteorological content element sample pair; and determining the relationship status sorting error of the meteorological content element sample in the meteorological content element sample pair based on the discrimination error comparison result and the evaluation weight comparison result.

[0116] In the application scenario of digital meteorological services, the big data management system plays a crucial role. This system is not only responsible for collecting, storing, and analyzing massive meteorological data, but also ensuring the validity and accuracy of these data. When implementing the relationship status sorting of meteorological content elements, the big data management system evaluates and processes the data based on a series of refined algorithms.

[0117] For example, when the big data management system processes meteorological content element samples, it will first conduct discrimination training for each meteorological content element sample, which will generate a discrimination training error of the meteorological content element. This error reflects the system's ability to correctly classify meteorological content elements. The smaller the error, the higher the classification accuracy.

[0118] Meanwhile, the system also assigns a relationship status evaluation weight to each meteorological content element sample. This weight is set based on the importance, representativeness of the sample, and its influence in the overall dataset. For example, some key meteorological elements, such as temperature, wind speed, etc., may be given higher weights because they are crucial for the accuracy of meteorological forecasts and services.

[0119] After determining the discrimination training error of meteorological content elements and the relationship status evaluation weight, the big data management system further calculates the relationship status sorting error of meteorological content element samples. This error is obtained by comparing the discrimination error and the evaluation weight. For example, the system first compares the discrimination errors between different samples to obtain a discrimination error comparison result; then, it compares the evaluation weights of different samples to obtain an evaluation weight comparison result.

[0120] The final relationship status sorting error is obtained based on the comprehensive consideration of the above two comparison results. This error reflects the overall ability of the system to accurately classify all meteorological content element samples after considering the weights. The formula can be expressed as: relationship status sorting error = Σ(discrimination error comparison result * evaluation weight comparison result). This formula is actually a weighted sum of the discrimination errors of each sample to obtain an overall error evaluation. In this way, the big data management system can more accurately understand its performance when processing different meteorological content elements, and thus optimize and improve targeted.

[0121] In this way, the application of the big data management system in digital meteorological services not only improves the processing efficiency and accuracy of meteorological data, but also provides more reliable data support for meteorological forecasts and services. This helps to improve the overall level of meteorological services and provide more accurate and timely meteorological information for the public and related professional fields.

[0122] In some possible embodiments, determining the relationship status sorting error of the meteorological content element samples in the meteorological content element samples based on the discrimination error comparison result and the evaluation weight comparison result includes: weighting the discrimination error comparison result and the evaluation weight comparison result to obtain a weighted comparison result; obtaining a target loss parameter; and determining the relationship status sorting error of the meteorological content element samples in the meteorological content element samples based on the target loss parameter and the weighted comparison result.

[0123] In the application scenario of digital meteorological services, when a big data management system processes meteorological data, it will encounter the problem of how to accurately evaluate and process meteorological content elements. This involves precisely sorting out the relationship status of meteorological content element samples to ensure the accuracy and effectiveness of the data. Now, how to determine the sorting error of the relationship status of meteorological content element samples based on the discriminant error comparison result and the evaluation weight comparison result will be introduced in detail.

[0124] First, the big data management system performs a weighting process on the previously mentioned discriminant error comparison result and evaluation weight comparison result. This weighting process assigns different weights according to the importance of each comparison result to obtain a weighted comparison result. The advantage of doing this is that it can more comprehensively reflect the differences and importance of different samples in the processing process.

[0125] Next, the system obtains a target loss parameter. This target loss parameter is a preset parameter used to measure the gap between the system's processing effect and the ideal state. It can help understand the gap between the current processing effect and the optimal effect, thus providing a direction for subsequent optimization.

[0126] After having the weighted comparison result and the target loss parameter, the big data management system can further determine the sorting error of the relationship status of meteorological content elements in the meteorological content element samples. Specifically, the system combines the weighted comparison result and the target loss parameter for calculation. This process can be understood as: the system comprehensively evaluates the gap between the currently processed meteorological content element samples and the ideal state by comparing the weighted discriminant error and evaluation weight and then combining the target loss parameter.

[0127] The calculation formula for this sorting error of the relationship status can be expressed as:

[0128] Sorting error of the relationship status = |Weighted comparison result - Target loss parameter|.

[0129] This formula describes the gap between the actual processing effect and the ideal state. When this error value is small, it indicates that the system's processing effect is close to the ideal state; otherwise, it indicates that there is still a large room for optimization.

[0130] In this way, the big data management system can more accurately evaluate its performance when processing meteorological content elements and make corresponding optimizations and adjustments according to the evaluation results. This not only improves the accuracy and efficiency of meteorological data processing but also provides more reliable data support for subsequent meteorological predictions and services. Thus, through the precise processing and optimization of the big data management system, more accurate and reliable meteorological data can be obtained, thereby improving the overall quality and level of digital meteorological services. This is of great significance for ensuring public life safety, guiding agricultural production, and supporting related scientific research activities, etc.

[0131] In an alternative embodiment, extracting the meteorological content element samples corresponding to the relationship state sorting error that meets the set requirements from the meteorological content element samples to obtain the target meteorological content element samples with mismatched discrimination training error and relationship state evaluation weight includes: if the relationship state priority of the meteorological content element sample is paired with the weighted comparison result, determining that the relationship state sorting error of the meteorological content element sample meets the set requirements; and using the meteorological content element sample corresponding to the relationship state sorting error that meets the set requirements as the target meteorological content element sample.

[0132] In the application scenario of digital meteorological services, when a big data management system processes and analyzes meteorological data, it needs to identify and extract key meteorological content element samples for more accurate meteorological prediction and services. In this process, the system not only has to process a large amount of data but also ensure the accuracy and effectiveness of the data. Now, a detailed introduction will be given on how to extract the meteorological content element samples corresponding to the relationship state sorting error that meets the set requirements from the meteorological content element samples.

[0133] First, the big data management system calculates the relationship state sorting error of each meteorological content element sample, which is obtained by comprehensively considering the discrimination training error and the relationship state evaluation weight. The discrimination training error reflects the system's ability to correctly classify meteorological content elements, while the relationship state evaluation weight represents the importance and influence of the element in the overall dataset.

[0134] When calculating the relationship state sorting error, the system combines the discrimination error comparison result and the evaluation weight comparison result and performs a weighted process to obtain a weighted comparison result. This weighted comparison result can more comprehensively reflect the differences and importance of different samples in the processing.

[0135] Next, the system determines whether the relationship state priority of each meteorological content element sample is paired with the weighted comparison result. The relationship state priority is set according to the importance and urgency of meteorological elements, which determines which elements should be given priority attention during the processing. If the relationship state priority of a certain meteorological content element sample matches the weighted comparison result, it means that the relationship state sorting error of this sample meets the set requirements. Once the meteorological content element samples corresponding to the relationship state sorting error that meets the set requirements are determined, the big data management system will use these samples as the target meteorological content element samples. These target meteorological content element samples are the data that need to be focused on and used in subsequent meteorological prediction and services.

[0136] In this way, the big data management system can more effectively screen out the data crucial for meteorological prediction and services, improving the pertinence and efficiency of data processing. At the same time, since the system extracts data based on the precise relationship state sorting error, the accuracy and reliability of the data are also greatly enhanced. Thus, through this embodiment of the big data management system, key meteorological content element samples can be more accurately identified and extracted, providing more accurate and effective data support for digital meteorological services. This not only helps improve the accuracy of meteorological prediction but also provides more timely and reliable meteorological information for the public and relevant professional fields, thus bringing more extensive social and economic benefits.

[0137] Under some preferred design ideas, the extraction of meteorological content elements from the dense interaction semantic samples to obtain each meteorological content element sample in the interaction information sample of the meteorological service platform includes: extracting meteorological content elements from the dense interaction semantic samples to obtain each initial meteorological content element sample in the interaction information sample of the meteorological service platform; extracting meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples.

[0138] In the application scenario of digital meteorological services, when the big data management system processes the interaction information of the meteorological service platform, it is necessary to accurately extract key meteorological content element samples. This process is crucial for providing accurate meteorological services. Now, how to obtain each meteorological content element sample by extracting meteorological content elements from the dense interaction semantic samples will be introduced in detail.

[0139] First, the big data management system conducts a preliminary extraction of meteorological content elements from the dense interaction semantic samples. This step is to obtain each initial meteorological content element sample in the interaction information of the meteorological service platform. These initial samples may contain a large amount of data, but not all of the information directly contributes to meteorological services, so further screening and refinement are needed.

[0140] After obtaining the initial meteorological content element samples, the system uses the initial meteorological element dense vectors of these samples for in-depth analysis. The dense vector is a mathematical representation that can capture semantic information and can help the system more accurately understand the meaning and context of each initial meteorological content element sample.

[0141] Next, the big data management system extracts truly valuable meteorological content element samples from the initial meteorological content element samples based on these dense vectors. This process is achieved by calculating the similarity between each sample and the predefined meteorological element features. The system selects those samples that highly match the key meteorological element features as the final meteorological content element samples.

[0142] To ensure the accuracy of extraction, the big data management system also verifies and optimizes the extracted meteorological content element samples. For example, the system can compare with other reliable data sources or use machine learning algorithms to correct and improve the extraction results.

[0143] Through the above steps, the big data management system can accurately extract key meteorological content element samples from the dense interaction semantic samples. These samples not only provide valuable data support for meteorological forecasting but also lay a solid foundation for subsequent services. In this way, through precise extraction of meteorological content elements, the big data management system can provide more accurate and personalized meteorological services. This can not only improve the accuracy and timeliness of meteorological forecasting but also help users better understand and cope with complex weather changes. At the same time, the optimized meteorological content element samples can also provide more reliable data support for relevant scientific research activities and promote the development of meteorological science.

[0144] In the following embodiments, extracting meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples includes: determining the initial relationship state evaluation weight of the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples; extracting the initial meteorological content element samples with the largest initial relationship state evaluation weight from the initial meteorological content element samples; determining the overlap coefficient between the content element coverage vector of the initial meteorological content element samples and the content element coverage vector of the initial meteorological content element samples with the largest initial relationship state evaluation weight; and if the overlap coefficient is less than the set coefficient, using the initial meteorological content element samples corresponding to the overlap coefficient as meteorological content element samples.

[0145] In the application scenario of digital meteorological services, the big data management system further processes the initial meteorological content element samples to extract key meteorological content element samples. This process is crucial for ensuring the accuracy and effectiveness of meteorological data. Next, how the system achieves this goal through a series of steps will be elaborated in detail.

[0146] First, the big data management system determines the initial relationship state evaluation weight of each sample based on the initial meteorological element dense vectors of the initial meteorological content element samples. The dense vector is a high-dimensional mathematical representation that can capture the complex relationships between data, and the relationship state evaluation weight reflects the importance and influence of each sample in the overall data set. The system assigns a weight value to each initial meteorological content element sample by calculating and analyzing these dense vectors.

[0147] Next, the system extracts the example with the largest initial relationship state evaluation weight from all the initial meteorological content element examples. This step is to find the most representative and influential examples in the dataset because they often contain the most critical meteorological information.

[0148] Then, the system determines the overlap coefficient between the content element coverage vectors of all the initial meteorological content element examples and the content element coverage vector of the example with the largest initial relationship state evaluation weight. The content element coverage vector is a mathematical quantity representing the content coverage range of an example, and the overlap coefficient measures the degree of content overlap between different examples.

[0149] Finally, if the overlap coefficient of a certain initial meteorological content element example is less than a preset setting coefficient, the system extracts this example as a key meteorological content element example. This step is to ensure that the extracted examples are unique and complementary in content, avoiding redundancy and repetition.

[0150] Through this series of steps, the big data management system can accurately extract key meteorological content element examples from the initial meteorological content element examples. These examples not only represent the core information in the dataset but also provide a solid data foundation for subsequent meteorological analysis and prediction. In this way, by accurately extracting key meteorological content element examples, the big data management system can more effectively utilize limited computing resources while improving the accuracy of meteorological prediction. In addition, this extraction method helps to identify and remove redundant information in the data, thereby further improving the data quality and the reliability of the analysis results.

[0151] In an exemplary embodiment, the dense interaction semantic mining of the meteorological service platform interaction information examples in the network learning example set to obtain the dense interaction semantic examples of the meteorological service platform interaction information includes: generating a first dense interaction semantic mining branch in the to-be-debugged structured data generation network, and performing dense interaction semantic mining on the meteorological service platform interaction information examples in the network learning example set to obtain the dense interaction semantic examples of the meteorological service platform interaction information.

[0152] Before determining the meteorological content element discrimination training error of the meteorological content element examples based on the meteorological element dense vectors of the meteorological content element examples and the relationship state description features of the meteorological service platform interaction information examples, it further includes: generating a second dense interaction semantic mining branch in the to-be-debugged structured data generation network, and performing dense interaction semantic mining on the meteorological content element examples to obtain the meteorological element dense vectors of the meteorological content element examples.

[0153] In the application scenario of digital meteorological services, when the big data management system processes the interaction information of the meteorological service platform, it will utilize a special network structure: the to-be-debugged structured data generation network, which contains multiple dense interaction semantic mining branches for deeply mining the deep meaning and associations of meteorological data.

[0154] In an exemplary embodiment, the big data management system first passes through a network learning example set, which contains a large number of example samples of meteorological service platform interaction information. The system conducts in-depth semantic mining on these interaction information samples through the first dense interaction semantic mining branch in the to-be-debugged structured data generation network. This process is to capture and extract the complex semantic relationships in the information, thereby generating dense interaction semantic samples of the meteorological service platform interaction information samples. These dense interaction semantic samples not only contain the core content of the original information but also reflect the internal connections and context relationships between the information.

[0155] Next, before preparing to calculate the discrimination training error of the meteorological content element samples, the big data management system further utilizes the second dense interaction semantic mining branch in the to-be-debugged structured data generation network. This time, the mining object of the system is the previously extracted meteorological content element samples. Through the in-depth mining of this branch, the system can generate meteorological element dense vectors of the meteorological content element samples. These dense vectors are high-dimensional mathematical representations that can accurately capture and express the key features and complex relationships of the meteorological content elements.

[0156] In summary, the big data management system conducts in-depth semantic mining on the meteorological service platform interaction information samples and the meteorological content element samples respectively through the two dense interaction semantic mining branches in the to-be-debugged structured data generation network. This not only provides a rich data basis for the subsequent calculation of the discrimination training error but also ensures the accuracy and effectiveness of meteorological data processing. In this way, the system can more accurately understand the internal meaning and relationships of meteorological data, and then provide more accurate and personalized meteorological services. Thus, through dense interaction semantic mining, the system can better capture the complex relationships and hidden information in the data, thereby improving the accuracy of meteorological prediction. At the same time, this processing method can also enhance the intelligent level of the system, enabling it to learn and adapt to the changing meteorological data more autonomously, bringing more efficient and reliable support for digital meteorological services.

[0157] Based on the above, after debugging the to-be-debugged structured data generation network according to the target discrimination training error to obtain a target structured data generation network, the method further includes: performing dense interaction semantic mining on the interaction information of the to-be-managed meteorological service platform through the target structured data generation network to obtain a dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform; determining local meteorological content elements in the interaction information of the to-be-managed meteorological service platform based on the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform; and refining the structured relationship state of the interaction information of the to-be-managed meteorological service platform based on the meteorological element dense vector of the local meteorological content elements and the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform to obtain the current structured relationship state of the interaction information of the to-be-managed meteorological service platform.

[0158] In the application scenario of digital meteorological services, after a series of data processing and model training, the big data management system has obtained an optimized target structured data generation network. Next, the system will use this well-trained network to conduct in-depth analysis and processing of the interaction information of the to-be-managed meteorological service platform.

[0159] First, the big data management system performs dense interaction semantic mining on the interaction information of the to-be-managed meteorological service platform through the target structured data generation network. The purpose of this step is to extract deep semantic features from complex interaction information and transform them into dense interaction semantic vectors. These vectors not only contain the core content of the interaction information but also reflect the internal connections and semantic relationships between the information.

[0160] Next, based on the mined dense interaction semantic vectors, the big data management system further determines the local meteorological content elements in the interaction information of the to-be-managed meteorological service platform. These elements are the specific meteorological contents in the interaction information, such as temperature, humidity, wind direction, etc., and they are the key to subsequent analysis and prediction.

[0161] Finally, the big data management system uses the meteorological element dense vector of the local meteorological content elements and the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform to refine the structured relationship state of the interaction information. This step is to reveal the relationship state between various elements in the interaction information, such as their correlation and dependence. Through the refined current structured relationship state, the system can more comprehensively understand the internal structure and law of meteorological data and provide more powerful support for subsequent meteorological prediction and services.

[0162] Applying the above embodiments enables the big data management system to more deeply understand and process the interaction information of the meteorological service platform, thereby providing more accurate and intelligent meteorological services. Through dense interaction semantic mining and structured relationship state refinement, the system can not only capture the core content and key elements of the interaction information, but also reveal the internal connections and laws between them, bringing higher efficiency and accuracy to digital meteorological services. In this way, through this processing method, the big data management system can more effectively utilize the interaction information of the meteorological service platform, improving the accuracy and timeliness of meteorological forecasts. At the same time, this method of in-depth analysis and structured refinement also helps to discover new meteorological laws and trends, providing more valuable data support for meteorological scientific research and business applications.

[0163] In some independent embodiments, after refining the structured relationship state of the interaction information of the to-be-managed meteorological service platform based on the dense vector of meteorological elements of the local meteorological content elements and the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform to obtain the current structured relationship state of the interaction information of the to-be-managed meteorological service platform, the method further includes: based on the current structured relationship state, performing structured storage on the interaction information of the to-be-managed meteorological service platform.

[0164] In some independent embodiments, the performing structured storage on the interaction information of the to-be-managed meteorological service platform based on the current structured relationship state includes: obtaining a set of session interaction relationship trees of the interaction information of the to-be-managed meteorological service platform based on the current structured relationship state, where the set of session interaction relationship trees includes X consecutive session interaction relationship trees, and X is an integer greater than or equal to 1; obtaining a set of interaction attribute relationship trees based on the set of session interaction relationship trees, where the set of interaction attribute relationship trees includes X consecutive interaction attribute relationship trees; obtaining a set of session interaction dense semantic networks based on the set of session interaction relationship trees through a first dense semantic embedding subnet included in the structured storage processing network, where the set of session interaction dense semantic networks includes X session interaction dense semantic networks; obtaining a set of interaction attribute dense semantic networks based on the set of interaction attribute relationship trees through a second dense semantic embedding subnet included in the structured storage processing network, where the set of interaction attribute dense semantic networks includes X interaction attribute dense semantic networks; obtaining a storage logic description feature corresponding to the set of session interaction relationship trees based on the set of session interaction dense semantic networks and the set of interaction attribute dense semantic networks through a storage logic construction subnet included in the structured storage processing network; determining a structured storage output feature of the set of session interaction relationship trees according to the storage logic description feature.

[0165] In the field of digital meteorological services, the big data management system plays a crucial role. It can not only efficiently process a vast amount of meteorological data but also perform intelligent storage and management based on the structured relationship status of the data. The following will elaborate in detail on an embodiment of how the big data management system performs structured storage on the interaction information of the meteorological service platform based on the current structured relationship status.

[0166] First, based on the current structured relationship status, the big data management system captures the interaction information on the meteorological service platform to be managed, and then generates a set of session interaction relationship trees. This set consists of consecutive X session interaction relationship trees, where X is an integer greater than or equal to 1. Each session interaction relationship tree details the interaction behaviors between users and the meteorological service platform, such as querying the weather, setting reminders, etc.

[0167] Next, according to these sets of session interaction relationship trees, the big data management system further extracts a set of interaction attribute relationship trees. This process aims to extract key attributes from complex interaction data, such as user ID, interaction time, interaction type, etc., for subsequent analysis.

[0168] After that, the big data management system uses a specially designed structured storage processing network to process this data. This network consists of several key parts: the first dense semantic embedding subnet, the second dense semantic embedding subnet, and the storage logic construction subnet.

[0169] The first dense semantic embedding subnet is responsible for processing the set of session interaction relationship trees and converting it into a set of session interaction dense semantic networks. This set contains X session interaction dense semantic networks, and each network deeply captures the semantic information of the interaction between users and the platform.

[0170] The second dense semantic embedding subnet processes the set of interaction attribute relationship trees to generate a set of interaction attribute dense semantic networks. These networks detail various attributes and their relationships during the interaction process.

[0171] Finally, the storage logic construction subnet will, based on the dense semantic network sets generated by the first two subnets, construct the storage logic description features corresponding to the set of session interaction relationship trees. These features not only reflect the interaction behaviors between users and the meteorological service platform but also reveal the logical relationships and patterns behind these behaviors.

[0172] Based on these storage logic description features, the big data management system finally determines the structured storage output features of the set of session interaction relationship trees. This means that the system can store and retrieve the interaction information on the meteorological service platform in an efficient and orderly manner, providing a solid foundation for subsequent data analysis, user behavior prediction, etc.

[0173] In this way, the big data management system not only improves the processing efficiency of meteorological data, but also enables the meteorological service platform to understand user needs more deeply, so as to provide more accurate and personalized services. At the same time, this structured storage method also helps to ensure the security and integrity of data, providing strong technical support for the digital transformation of the meteorological service industry.

[0174] In some other independently implementable embodiments, based on the session interaction dense semantic web set and the interaction attribute dense semantic web set, obtaining the storage logic description features corresponding to the session interaction relationship tree set through the storage logic construction subnet included in the structured storage processing network includes: based on the session interaction dense semantic web set, obtaining X first session vector knowledges through the first multi-head feature focusing module included in the structured storage processing network, where each first session vector knowledge corresponds to a session interaction dense semantic web; based on the interaction attribute dense semantic web set, obtaining X second session vector knowledges through the second multi-head feature focusing module included in the structured storage processing network, where each second session vector knowledge corresponds to an interaction attribute dense semantic web; performing integration processing on the X first session vector knowledges and the X second session vector knowledges to obtain X target session vector knowledges, where each target session vector knowledge includes a first session vector knowledge and a second session vector knowledge; based on the X target session vector knowledges, obtaining the storage logic description features corresponding to the session interaction relationship tree set through the storage logic construction subnet included in the structured storage processing network.

[0175] Based on this embodiment, when the big data management system processes the session interaction information in digital meteorological services, it will take a series of refined steps to extract the storage logic description features. This process mainly relies on the structured storage processing network, especially the storage logic construction subnet therein, and the cooperating multi-head feature focusing module.

[0176] First, the big data management system utilizes the session interaction dense semantic web set, which is the key information extracted from the user sessions of the meteorological service platform through previous steps. This information is sent to the first multi-head feature focusing module in the structured storage processing network. The role of this module is to extract the most critical features from each session interaction dense semantic web to generate the so-called "first session vector knowledges". These vector knowledges are highly condensed information representations, each corresponding to a specific session interaction content. In this way, the system can efficiently process a large amount of user session data while retaining its core information.

[0177] Meanwhile, the big data management system also performs similar processing on the collection of interaction-attribute dense semantic webs. These webs contain various attribute information of the user's interaction with the meteorological service platform. This information is fed into the second multi-head feature focusing module, which extracts "second session vector knowledge" from it. Similar to the first session vector knowledge, these vector knowledges are also highly refined information units, each representing the comprehensive features of an interaction attribute.

[0178] Next, the system integrates the first session vector knowledge and the second session vector knowledge. The purpose of this step is to fuse the specific content of the user session and the related attribute information together to form a more comprehensive and richer data representation. Through this integration processing, the system can generate "target session vector knowledge", which contains both the detailed content of the user session and the key interaction attributes.

[0179] Finally, based on these target session vector knowledge, the big data management system generates the storage logic description features corresponding to the set of session interaction relationship trees through a storage logic construction subnet. These features not only reflect the actual interaction situation between the user and the meteorological service platform, but also reveal the logical structure and correlation relationships behind these interactions. This is crucial for subsequent data storage, retrieval, and analysis, as it provides an efficient and orderly way to manage and utilize these valuable information resources.

[0180] In this way, the big data management system can understand the user's needs and behavior patterns more deeply, thus providing more accurate and personalized support for digital meteorological services. At the same time, this processing method also greatly improves the efficiency of data storage and management, laying a solid foundation for the sustainable development of the meteorological service industry. Generally speaking, this technical solution not only improves the intelligent level of the meteorological service platform, but also brings a better quality and more convenient service experience to users.

[0181] In some other independent embodiments, based on the session interaction dense semantic web set, X first session vector knowledge is obtained through the first multi-head feature focusing module included in the structured storage processing network, including: for each session interaction dense semantic web in the session interaction dense semantic web set, a first local encoded dense semantic web is obtained through the local encoding layer included in the first multi-head feature focusing module, where the first multi-head feature focusing module belongs to the structured storage processing network; for each session interaction dense semantic web in the session interaction dense semantic web set, a first global sampled dense semantic web is obtained through the global sampling layer included in the first multi-head feature focusing module; for each session interaction dense semantic web in the session interaction dense semantic web set, a first cross dense semantic web is obtained through the residual layer included in the first multi-head feature focusing module based on the first local encoded dense semantic web and the first global sampled dense semantic web; for each session interaction dense semantic web in the session interaction dense semantic web set, a first session vector knowledge is obtained through the first global sampling layer included in the first multi-head feature focusing module based on the first cross dense semantic web and the session interaction dense semantic web.

[0182] In some other independent embodiments, when the big data management system processes the session interaction data in digital meteorological services, it will use a complex structured storage processing network to extract key information. An important part of this network is the first multi-head feature focusing module, which is responsible for extracting the first session vector knowledge from the session interaction dense semantic web set.

[0183] First, the big data management system processes each network in the session interaction dense semantic web set through the local encoding layer in the first multi-head feature focusing module. The role of this layer is to encode the local features of each session interaction dense semantic web, capture the detailed information therein, and generate the so-called first local encoded dense semantic web. This process helps the system better understand the specific interaction content between the user and the meteorological service platform.

[0184] Next, the system continues to process each session interaction dense semantic web using the global sampling layer in the first multi-head feature focusing module. The purpose of the global sampling layer is to grasp the features of each network as a whole, extract the global information, and generate the first global sampled dense semantic web. This step helps the system master the overall trend and pattern of user interactions.

[0185] Then, based on the first local-encoded dense semantic network and the first global-sampled dense semantic network generated previously, the system performs further processing through the residual layer in the first multi-head feature focusing module. The role of the residual layer is to effectively fuse local features and global features to generate the first cross-dense semantic network. This fusion method can retain more original information while enhancing the expressive power of features.

[0186] Finally, the system once again uses the global sampling layer (referred to here as the first global sampling layer to distinguish it from the previous global sampling layer) to jointly process the first cross-dense semantic network and the original session interaction dense semantic network. The purpose of this step is to extract the most crucial information from the fused features to generate the first session vector knowledge. These vector knowledge not only contain the local details of the user session but also reflect the global interaction patterns, providing strong support for subsequent data storage and analysis.

[0187] In this way, the big data management system can more accurately capture and analyze the interaction information between users and the digital meteorological service platform. This not only helps improve the personalization and intelligence levels of meteorological services but also provides a solid foundation for the in-depth mining and effective utilization of meteorological data. Generally speaking, this technical solution significantly improves the efficiency and accuracy of the big data management system in processing complex interaction data, injecting new vitality into the development of digital meteorological services.

[0188] Furthermore, Figure 2 FIG. is a schematic structural diagram of a big data management system 200 provided by an embodiment of the present application. As Figure 2 shown, the big data management system 200 includes a processor 210, and the processor 210 can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0189] Optionally, as Figure 2 shown, the big data management system 200 may further include a memory 230. Among them, the processor 210 can call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.

[0190] Among them, the memory 230 can be a separate device independent of the processor 210 or integrated in the processor 210.

[0191] Optionally, as Figure 2 shown, the big data management system 200 may further include a transceiver 220, and the processor 210 can control the transceiver 220 to interact with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices.

[0192] Optionally, the big data management system 200 may implement the corresponding processes of the storage engine or components (such as a processing module) in the storage engine or the device on which the storage engine is deployed in each method of the embodiments of the present application. For the sake of brevity, details are not described herein again.

[0193] It should be understood that the processor in the embodiments of the present application may be an integrated circuit chip with signal processing capabilities. In the implementation process, the steps of the above method embodiments may be completed by the integrated logic circuit in the hardware of the processor or instructions in software form. The above processor may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application may be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0194] It can be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include but not be limited to these and any other suitable types of memory.

[0195] It should be understood that the above-mentioned memory is by way of example but not limitation. For example, the memory in the embodiments of the present application can also be a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDR SDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchlink dynamic random access memory (SLDRAM), and a direct rambus random access memory (DR RAM), etc. That is to say, the memory in the embodiments of the present application is intended to include but not be limited to these and any other suitable types of memory.

[0196] On the basis described above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0197] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application.

Claims

1. A data management method for a meteorological service platform based on artificial intelligence, characterized in that, Applied to a big data management system, the data management method of the artificial intelligence-based meteorological service platform includes: Obtain a network learning example set for the to-be-debugged structured data generation network, and perform dense interaction semantic mining on the meteorological service platform interaction information samples in the network learning example set to obtain dense interaction semantic samples of the meteorological service platform interaction information samples; Extract meteorological content element samples from the dense interaction semantic samples to obtain various meteorological content element samples in the meteorological service platform interaction information samples, and determine the relationship state evaluation weight of the meteorological content element samples based on the dense vectors of meteorological elements of the meteorological content element samples; among them, the dense vector of meteorological elements is a compact numerical representation of the meteorological content element sample, which fuses multiple meteorological elements into a high-dimensional vector, and each dimension represents a corresponding meteorological feature; the relationship state evaluation weight is the relative importance occupied by each element when evaluating the relationship or state between meteorological content element samples; Determine the meteorological content element discrimination training error of the meteorological content element samples based on the dense vectors of meteorological elements of the meteorological content element samples and the relationship state description features of the meteorological service platform interaction information samples; Extract target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element samples, and determine the target discrimination training error of the meteorological service platform interaction information samples based on the dense vectors of meteorological elements of the target meteorological content element samples and the dense interaction semantic samples; Debug the to-be-debugged structured data generation network based on the target discrimination training error to obtain a target structured data generation network, so as to refine the structured relationship state of the meteorological service platform interaction information through the target structured data generation network; among them, the refined structured relationship state is used to realize the construction of structured data of the meteorological service platform interaction information.

2. The data management method of the meteorological service platform based on artificial intelligence according to claim 1, wherein, Extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element samples includes: Extract binary groups of meteorological content element samples from the meteorological content element samples; Based on the meteorological content element discrimination training errors and relationship state evaluation weights of the meteorological content element samples in the binary group of meteorological content element samples, extract target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element samples.

3. The data management method of the weather service platform based on artificial intelligence according to claim 2, characterized in that, The extracting target meteorological content element samples with mismatched meteorological content element discrimination training errors and relationship state evaluation weights from the meteorological content element samples based on the meteorological content element discrimination training errors and relationship state evaluation weights of the meteorological content element samples in the binary group of meteorological content element samples includes: Determine the first comparison result between the relationship state evaluation weights of the meteorological content element samples in the binary group of meteorological content element samples based on the meteorological content element discrimination training errors of the meteorological content element samples in the binary group of meteorological content element samples; Determine a second comparison result among the relationship status evaluation weights of the meteorological content element examples in the meteorological content element example binary group based on the relationship status evaluation weights of the meteorological content element examples in the meteorological content element example binary group; If the first comparison result does not match the second comparison result, determine the target meteorological content element example in the meteorological content element example binary group where the meteorological content element discrimination training error and the relationship status evaluation weight do not match.

4. The data management method of the artificial intelligence-based meteorological service platform according to claim 2, wherein The extracting, from the meteorological content element examples, of the target meteorological content element examples where the meteorological content element discrimination training error and the relationship status evaluation weight do not match based on the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element examples in the meteorological content element example binary group includes: Determine the relationship status sorting error of the meteorological content element examples in the meteorological content element examples based on the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element examples in the meteorological content element example binary group; Extract the meteorological content element examples corresponding to the relationship status sorting error that meets the set requirements from the meteorological content element examples to obtain the target meteorological content element examples where the meteorological content element discrimination training error and the relationship status evaluation weight do not match.

5. The data management method of the meteorological service platform based on artificial intelligence according to claim 4, wherein, The determining, based on the meteorological content element discrimination training error and the relationship status evaluation weight of the meteorological content element examples in the meteorological content element example binary group, of the relationship status sorting error of the meteorological content element examples in the meteorological content element examples includes: Determine the discrimination error comparison result between the meteorological content element examples in the meteorological content element example binary group based on the meteorological content element discrimination training error of the meteorological content element examples in the meteorological content element example binary group; Determine the evaluation weight comparison result between the meteorological content element examples in the meteorological content element example binary group based on the relationship status evaluation weight of the meteorological content element examples in the meteorological content element example binary group; Determine the relationship status sorting error of the meteorological content element examples in the meteorological content element examples based on the discrimination error comparison result and the evaluation weight comparison result.

6. The data management method of the meteorological service platform based on artificial intelligence according to claim 5, wherein The determining, based on the discrimination error comparison result and the evaluation weight comparison result, of the relationship status sorting error of the meteorological content element examples in the meteorological content element examples includes: Perform weighting on the discrimination error comparison result and the evaluation weight comparison result to obtain a weighted comparison result; Obtain a target loss parameter, and determine the relationship status sorting error of the meteorological content element examples in the meteorological content element examples based on the target loss parameter and the weighted comparison result; Wherein, the extracting, from the meteorological content element examples, of the meteorological content element examples corresponding to the relationship status sorting error that meets the set requirements to obtain the target meteorological content element examples where the discrimination training error and the relationship status evaluation weight do not match includes: If the relationship status priority of the meteorological content element sample is paired with the weighted comparison result, it is determined that the relationship status sorting error of the meteorological content element sample meets the set requirements; Use the meteorological content element sample corresponding to the relationship status sorting error that meets the set requirements as the target meteorological content element sample.

7. The data management method of the artificial intelligence-based meteorological service platform according to claim 1, wherein The extraction of meteorological content elements from the dense interaction semantic sample to obtain each meteorological content element sample in the interaction information sample of the meteorological service platform includes: Extract each initial meteorological content element sample in the interaction information sample of the meteorological service platform by extracting meteorological content elements from the dense interaction semantic sample; Extract meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples; Among them, the extraction of meteorological content element samples from the initial meteorological content element samples based on the initial meteorological element dense vectors of the initial meteorological content element samples includes: Determine the initial relationship status evaluation weight of the initial meteorological content element sample based on the initial meteorological element dense vector of the initial meteorological content element sample; Extract the initial meteorological content element sample with the largest initial relationship status evaluation weight from the initial meteorological content element samples; Determine the overlap coefficient between the content element coverage vector of the initial meteorological content element sample and the content element coverage vector of the initial meteorological content element sample with the largest initial relationship status evaluation weight; If the overlap coefficient is less than the set coefficient, use the initial meteorological content element sample corresponding to the overlap coefficient as the meteorological content element sample.

8. The data management method of the meteorological service platform based on artificial intelligence according to claim 1, wherein, The dense interaction semantic mining of the interaction information sample of the meteorological service platform in the network learning example set to obtain the dense interaction semantic sample of the interaction information sample of the meteorological service platform includes: generating a first dense interaction semantic mining branch in the to-be-debugged structured data generation network, and performing dense interaction semantic mining on the interaction information sample of the meteorological service platform in the network learning example set to obtain the dense interaction semantic sample of the interaction information sample of the meteorological service platform; Before determining the meteorological content element discrimination training error of the meteorological content element sample based on the meteorological element dense vector of the meteorological content element sample and the relationship status description feature of the interaction information sample of the meteorological service platform, it further includes: generating a second dense interaction semantic mining branch in the to-be-debugged structured data generation network, and performing dense interaction semantic mining on the meteorological content element sample to obtain the meteorological element dense vector of the meteorological content element sample.

9. The data management method of the artificial intelligence-based meteorological service platform according to claim 1, characterized in that After debugging the to-be-debugged structured data generation network based on the target discrimination training error to obtain the target structured data generation network, the method further includes: Performing dense interaction semantic mining on the interaction information of the to-be-managed meteorological service platform through the target structured data generation network to obtain the dense interaction semantic vector of the interaction information of the to-be-managed meteorological service platform; Determine the local meteorological content elements in the interaction information of the meteorological service platform to be managed based on the dense interaction semantic vectors of the interaction information of the meteorological service platform to be managed; Based on the meteorological element dense vectors of the local meteorological content elements and the dense interaction semantic vectors of the interaction information of the meteorological service platform to be managed, refine the structured relationship state of the interaction information of the meteorological service platform to be managed, and obtain the current structured relationship state of the interaction information of the meteorological service platform to be managed.

10. A big data management system, characterized in that, It includes at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1-9.

Images